The Lewis & Short dictionary contains useful information about this:
- fĕmur, ŏris or ĭnis (acc. to a nom. ‡ femen, mentioned only by Prisc. p. 701 P. and Serv. Verg. A. 10, 344; 778; nom. femus, Ap. M. 8, p. 216, 15; cf.: μῆρος, femus, Gloss. Lab.; dat. femori; femini only Plin. 28, 15, 61, § 217; abl. usually femore, but femine, Plaut. Mil. 2, 2, 48; Verg. A. 10, 788; plur. femora or femina; dat. feminibus, rarely femoribus)
The authors lived in the 19th century and might have partially relied on intuition for their judgement (even though they probably had concordances, these may not have been easy to use or fully available to them).
Checking
femore vs.
femine on Packhum, which contains recent or rather recent editions of most of the well-known classical works plus some late-imperial texts, it seems they made an alright call in
femine vs.
femore, although as far as Packhum goes,
femine appears about twice as often (n=19) as
femore (n=9). Please note the numbers are rather low in both cases, and that this word has precisely the kind of situation where manuscripts may not reliably relay the original text: both femine and femore fit meters equally in poetry, and the same goes for the other cases.
It is true
feminibus is far more common (n=23) than
femoribus (n=2, once Celsus and once Apuleius).
Feminī only appears in Pliny the Elder among the well-known classical works in Packhum's editions, just as in Lewis & Short's.
Checking
femora vs.
femina would involve a lot of work because of the homography between our word
fĕmina 'thigh' and
fēmina/fēminā 'woman', the latter being an extremely common word. It is left as an exercise to the reader (please inform us if you do carry it out). Checking
femoris vs.
feminis is a problem for the same reason.
However, Packhum only contains
femora in a couple practical texts (Celsus, Apicius), even though it should be somewhat common in reference to both thighs. This makes me think
femina might be a lot more common for the nominative/accusative plural. I also suspect
feminis might be more common as once again it's mostly just Celsus and Apuleius who use
femoris.
The authors do not mention this, but
feminum also seems to be a lot more common (n=12, careful with Varro) than
femorum (n=2, once Ovid and once Lucan).
So, this means that, yes, this word has a mixed declension.
nom./acc. sg.: femur (only in Apuleius: femus, also mentioned by Servius but with disapproval; also once in Ampelius: femen, mentioned by Servius with grudging approval)
gen. sg.: feminis (?
exact difference with femoris unknown, but femoris only in Celsus/Apuleius, also mentioned by Servius and with surprising approval)
dat. sg.: femorī (only in Pliny the Elder: feminī)
abl. sg.: femine or femore (femine is perhaps a bit more common)
nom./acc. pl.: femina (?
exact difference with femora unknown, but femora only once in Celsus/Apicius)
gen. pl.: feminum (rare but in Ovid/Lucan: femorum)
dat./abl. pl.: feminibus (rare but in Apuleius/Celsus: femoribus)
Caveat: again, if we rely on the editions Packhum has. I notice that De Vaan in his Latin etymological dictionary used a concordance of Cicero's works where
femoris appears. Differences like this could matter a lot.
Miscellaneous information that's irrelevant to your question but that I personally find fascinating:
The ancient, unattested Proto-Indo-European language, which Latin/Greek/English/Russian/Persian/Sanskrit and various other languages are the descendants of, had a category of nouns with an odd declension that ended in -r in the nominative/accusative/vocative cases, but -n in the genitive/dative/ablative/locative/instrumental cases.
As time passed, most Indo-European languages regularized these nouns by having only -r or only -n throughout. For example, Proto-Indo-European had peth2r pth2ens for 'feather' ~ 'wing', but then Latin shows up keeping only the -n (*petna > penna) while Greek kept only the -r (πτερόν pteron).
However, Latin still had a few nouns maintaining or kind of maintaining the old pattern: femur feminis/femoris 'thigh' (femoris is the new regularized form), iecur iecinoris/iecoris 'liver', iter itineris 'path, journey'.
Interestingly, although iecur and iter have Indo-European etymologies (with cognates in other languages), De Vaan in his etymological dictionary of Latin is unable to link femur with anything. I wonder if this plays a role in its varying declension with the popular nom./acc. sg. femus/femen forms lying around. Could it be that femur was borrowed from a non-Indo-European language into the language and was somehow fit into the -r/-n pattern?