beside what you've mentioned i think its mainly used in conjunction with subentity materials allowing you have multiple materials on the same mesh with the same skeleton( e.g change dress color etc... )
rambling:
after messing around mainly with mesh import/export/manipulation over the last year, i must admit that i'm starting to feel that all this submesh/subentity over-complicates stuff while contributing very little
i'm not a cg artist so my knowledge is very limited, but most free models i've downloaded from the web don't have submeshes, so i tend to think its mainly used as an optimization technique in game-engines, being ignored by the rest of the cg community who does'nt care much about millisecond scale optimizations
that being said...ogre still strives to be super-optimized and although submeshes over complicate simple stuff like mesh manipulation they do have their advantages, maybe offering some kind of middle-ground..
if you decide to split your sub-meshed model to many singular meshes you would need all singular meshes to share/sync with the same skeleton somehow( duplicate bones? ) and you'll lose some optimization( not sure how big/important that is, espicailly for ogre 2.1 )
and if you decide to just merge all sub-meshes to a singular mesh you would lose the subEntity-material feature
so i guess the question comes down to whether you need this advantages in your engine
off course, you can alway's implement your own animation-pipeline but that is troublesome...
interesting topic...cheers...
edit: another big optimization benefit i can think of is with hardware pose/blendshapes animation( e.g facial animation on face submesh )