The formula for limonite on the limonite page http://www.mindat.org/min-2402.html is given as FeO(OH)·nH2O
Yet on the same page is stated "Currently used for unidentified massive hydroxides and oxides of iron, with no visible crystals, and a yellow-brown streak. 'Limonite' is most commonly the mineral species goethite, but can also consist of varying proportions of maghemite, hematite, lepidocrocite, hisingerite, pitticite, jarosite group species, etc. " and further down is stated "While a "generic" formula is often given as stated here, "limonite" is really quite variable in composition - both in its levels of hydration and hydroxylation and in its cationic content. There really is no formual for it..."
First - need to fix the spelling error (formual ==> formula)
Second - giving a specific formula contradicts the statements, perhaps the formula should be changed to "impure hydrated Fe oxides and hydroxides" 
Thanks for noting this inconsistency - formula and "comments" text now deleted. 
I don't believe that limonite able to contain maghemite and hematite in its composition. 
Under tropical weathering conditions it often contains maghemite. The word doesn't refer to a specific substance (at least not in english), but just "fine grained, massive or powdery iron oxides of undefined composition". In European literature it seems to be sometimes treated as an actual mineral itself. Dear Alfredo,
temperature of dehydratation of pure goethite is 400o C, impure goethite may to dehidrate under 250o C with formation hematite
Under slow dehydration at 95o C lepidocrocite transforms into hematite, maghemite grow from lepidocrocite during fast heating at 170-400o C under atmosphere pressure.
I can to change my words, I don't think, that mixture of iron oxides containing maghemite and hematite can to be named limonite.
For example, maghemite in my collection is represented by powdery product of natural burning of peat from Belorussia - it is obviously not limonite.
So call "desert tan" over gravel in Gobi desert also isn't limonite. Limonite is product of precepitation from aqueous solutions, it unable to contain anhydrous minerals. Pavel: "mature" limonite/goethite often contains (secondary) hematite, especially in arid and/or +/- tropical environments.
As an example, see ref. below.
Maghemite: http://www.mindat.org/photo-343803.html
Morris, R. C. (1980): A textural and mineralogical study of the relationship of iron ore to banded iron-formation in the Hamersley iron province of Western Australia. Economic Geology, 75, 184-209.
Microscope studies of a wide range of ores and fresh, oxidized, or partly iron-enriched, banded iron-formation (BIF) have shown a consistent sequence of alteration of the primary constituents, chert, magnetite with or without hematite, carbonates, and silicates, leading to the formation of martite-(hematite)-goethite ores which form the bulk of the reserves of the Hamersley Iron Province. Primary hematite survives with little change; magnetite oxidizes to hematite (martite) or in part to lacunar phases between magnetite and maghemite (kenomagnetite), which in turn commonly is altered by hydration to goethite. Subsequent leaching of this goethite in preference to martite leads to a variety of skeletal textures which may help identify the original magnetite, despite destruction of martite lamellae by recrystallization. Chert, carbonates, and silicates are either replaced by goethite or leached from the system, resulting in thinning of the orebodies with respect to their parent banded iron-formations. Preservation of macroscopic features, such as mesobanding, and microscopic features, such as pseudomorphs, is compelling evidence of supergene enrichment which, together with residual upgrading by leaching of these iron-formations, produces ores which typically assay 62 percent Fe, ignition loss of 4 to 6 percent, and >0.07 percent P. With increasing maturity, part of the goethite has converted to fine-grained hematite in some of these martite-(hematite)-goethite ores. Though overall the secondary hematite is quantitatively unimportant, some deposits may be significantly upgraded by this process, and it is this variability that is implied by the use of brackets in the classification.A second ore type, characteristic of the Tom Price and Whaleback mines, consists essentially of residual primary components, martite and hematite, with abundant secondary hematite, most of it platy but some anhedral. These martite-hematite ores appear to have formed by a low-temperature process in which some of the goethite or its precursor, in ores of the martite-(hematite)-goethite type, has converted to platy hematite in contrast to the fine-grained and often cryptocrystalline material of the normal ores. This partial transformation has resulted in increased permeability of the deposits which in turn has allowed leaching of the remnant goethite. The Paraburdoo 4E mine represents an arrested stage of development in which a large proportion of this unconverted goethite is still present, whereas the Tom Price and Whaleback deposits represent the most mature ores in the province. In these, leaching has removed the bulk of the goethite, leading to typical assay values of >64 percent Fe, 1 percent loss on ignition, and <0.05 percent P. The supergene ores of the province can be classified in terms of a maturity scale based on the type and amount of secondary hematite replacing goethite. Thus martite-goethite ores without significant secondary hematite are considered to be the least mature. Examples of these have been found in a limited survey of the Marra Mamba Iron Formation ores. The Brockman Iron Formation deposits at Koodaideri, the Brockman syncline, and Metawandy provide examples of more mature ores with increased secondary hematite and are classified as martite-(hematite)-goethite ores. Ancient deposits (early to middle Proterozoic) with abundant secondary platy hematite but with significant residual goethite are classified as the Paraburdoo type. Substantial leaching of this goethite leads to the most mature ores of the province, the martite-hematite deposits of the Tom Price-Whaleback type. Dear Uwe,
you of course right - goethite of course able to transform into hematite or maghemite into metamorphic conditions.
But why you decide, that these maghemite pellets from Cobar has some relation to limonite?
Limonites sensu stricto are products hydratation and decomposition of FeSO4 in water solutions. Sourses of this FeSO4 are oxidation of different sulphides (not only iron sulphides - limonite able to form during oxidation of Fe-bearing sphalerites for example or from chalcopyrite).
Another widespread sourse of limonite is decarbonatization of siderite, ferroan dolomite etc.
The third source is primary precepitation from waters (sea, lake or swamp) - in this case Fe-sulphates are important intermediate (transport) member.
The fourth sourse of limonite is hydratation of Fe-bearing silicates during formation of laterite crusts of weathering.
In any case limonites are very young formations - with age not more than first MA. Banded iron-formations of Archaean age is bad example and not the best substrate for limonite formation. But it able to form even here along shear zones within BIF bodyes (which usually still contains some reactable iron sulphides and carbonates).
Limonite sensu stricto very often with coming of time recrystallized into crystalline agregates of iron hydroxides with admixtures of crystalline sulphates (jarosites) and finish to be limonites.
On the photo you may to see product of such transformation limonite layer into gorthite-lepidocrocite-(+/- natrojarosite) aggregate (concretion). The sample was collected on boundary of clay/sand bed rocks with layer of coal from Tula oblast' in Russia.
Here limonite is almost invisible because it is on back side of the sample.
Here is visible red lepidocrocite layer within goethite matrix. Minor preserved primary yellow powdery limonite is visible at the upper part of the photo.
Minor yellow powdery aggregates of secondary (later formed) limonite is visible in cavities.
This sample as whole can't to be named limonite, it contains some limonite, and it was formed after limonite recrystallization. Initial limonite in this case was formed by hydratation and decomposition of FeSO4 solutions, arised during oxidation of pyrite in coals. Age of this aggregate is after glacial (<25000-30000 years). 
Limonite is usually considered a "garbage" term for undetermined iron oxides/hydroxides which may or may not be crystalline. The iron analog of "wad". 
Yes it seems we vary in our usage of the term limonite, many of us would call Pavels whole soecimen a limonite concretion, not just the yellow powder, when we suspect a mixture.
In Australia we frequently have abundant ironstone pebbles in or covering our soils, aka pisolitic limonite. These form in the soils, possibly from weathering as old as Tertiary. Mostly they are goethite, sometimes lepidocrocite and sometimes amorphous material. But haematite and maghemite are also very common, and these mostly form from heating by wildfires. Limonite seem seems a convenient term to describe the whole assemblage. Whole this concretion, at least its solid brown part is pure goethite with some (quite negligible) excess of adsorbed water. Molecular water content in it according to thermal analysis is equal to *0.06H2O. In other words it is pure goethite - α-Fe3+O(OH) more that on 99 mas.%.
This concretion composed by at least 4 well individualized components - yellow earthy limonite (of different time of formation), brown crystallized pure goethite, red well crystallized lepidocrocite and velvet-black encrustation of inner hole (unidentified iron oxyde-hydroxyde). This isn't mixture, it is aggregate of 4 different individual components well visible by unequipped eye. Similar to as granite isn't mixture of quartz, microcline, albite and biotite.
Should we enter in the database lherzolite as mixture of olivine, clinopyroxene and chromite? Pavel, you are illustrating very well the different uses of the word limonite in different parts of the world. Among english-speaking tribes, "limonite" is no longer considered to be a discrete, definable substance, so it can't be approved by the IMA as a mineral. It has variable composition and structure. So that yellow earthy component that you call 'limonite", what is it exactly? Is it amorphous? or does it have a structure? Does it jave a fixed composition? We don't understand what you mean when you call it "limonite", because to the rest of us "limonite" just means "uncharacterised iron oxides and/or hydroxides". If there are some limonites which can be characterised, then you should describe them and submit them for consideration as new species. For me this http://www.mindat.org/photo-176578.html is typical limonite up to moment when it will be proven, that this is real schwertmannite or at least schwertmannite-bearing mixture (= limonite). But you are ready to naming this limonite schwertmannite, while obvious goethite naming limonite. Where is logic!? How about the following text to emphasize the imprecise nature of the term "limonite":
"Limonite" is a generic name given to otherwise unidentified massive hydroxides and oxides of iron with no visible crystals and a yellow-brown streak. Most materials described as "limonite" consist primarily of the mineral species goethite, but they may also contain one or more of maghemite, hematite, lepidocrocite, hisingerite, pitticite, jarosite group species, etc. Ed, that sounds a fairly good definition to me!
Limonite is basically a field term for imprecisely characterised material as Ed described. Similarly "granite" is commonly used as a field term for coarse grained feldspathic rocks ( though we should really use granitoid or granitic rock). On careful study in the lab we can classify them as maybe hornblende granodiorite, biotite monzonite or whatever, but till then people have a rough idea of what we are talking about.
Pavel, of course granite and lherzolite are mxtures of those minerals, but geologists somewhat arbitrarily divide up these mixtures and give them names based on varying mineral proportions and textures. Humans are always trying to breakdown complex natural systems and categorise them into little boxes, with varying success! Nature tends to thumb its nose at our attempt at categorisation. 
I have to say my view of limonite is closer to Pavel's than others.
Trying to call a whole concretion 'limonite' is a little bit lazy. it's a rock rather than a mineral anyway. an iron concretion that contains at least two mineral species.
The formula listed for limonite is one that's been used for years and years as an approximation.
Removing the formula might satisfy some purists here, but it does now remove limonite from any of the chemical searches. Not a good idea.
I'll put a (different) formula back in. Sorry guys Dear Ralph,
this stone also is mixture of unidentified iron oxydes and hydroxides, but it obviously isn't limonite (with exeption of brownish earthy crusts on its back side).
Its locality is Mushugai Khuduk in Gobi. Initially it was titanomagnetite from very specific neck-like body filled with celestine-anhydrite-apatite-magnetite magmatic melt. With time this titanomagnetite decomposed with formation of magnetite-ilmenite-maghemite-hematite-goethite mixture.
This is fragment of such neck, polished by wind&sand from all sides. Jolyon, if you are calling limonite a rock, which is not far wrong, then it shouldnt have a formula, you dont have formulae for rocks even if they are monominerallic. You need to realise its much like the term clay, it can be a rock or can be a mineral, but doesnt have a formulae as such. We need scientific credibility here, not trying to enable chemical searches for mixtures and rocks. Im not sure what is lazy about calling such a concretion limonite, its accurate and just calling it iron sounds even worse!
Pavel, interesting rock, but most geologists would still call it limonite as a field term, pending analyses. Or maybe ironstone. Perhaps you have other names in Russian for such things? No, I was saying that the iron concretions are rocks, not the limonite :) Same case like psilomelane :-) Yes Jakub, limonite is a term like psilomelane, wad, ochre, clay, etc that inhabits a limbo land between rocks and minerals, perhaps best called mixtures but should not be labelled minerals and given formal formulae etc, its misleading until they get analysed. Then some will be minerals, some mixtures and some rocks. Pavels examples are clearly monomineralic and well crystalline so the formal names apply, the same as much limonite can be labelled goethite when analysed or the crystal structure is visible. And if any brave mineralogist wants to tackle 'varlamoffite' some day, that's a real mess too
