It could very well be unclosed spans...but more likely unOPENed spans. I tend to find all the opened ones...not so much the closing tags. Wait..that can't be right, Sigil usually finds and nukes those. Must be unclosed ones, then, although the wondrous NoteTab usually makes unclosed tags obvious, even to dunderheads like yours truly.
Charleski: I don't really choose to "generate" epub files from Word; I receive Word files from clients (ether direct or as Abbyy output from scanners). I tried Atlantis--briefly, admittedly--and wasn't blown away. Really, cleaning up Word files, as long as you have a method to tag the italicization, is easy as hell--you just click "Clear Formatting" and bobs-yer-uncle.
But I did NOT use BD for these last two, electing insted to output the html and clean it myself, because I get pissed-off at having to scroll through the 2000 lines of ss that BD puts in the exported html file--even "filtered," since BD seems to feel some compulsion to embed every font in the universe, for reasons that completely elude me--in order to cut it out of the html file.
I was able to get BD to export the html once, without nine bazillion lines of ss, but I've never been able to replicate it, so I was trying to eliminate that step and regex all the crapola out of the file. Clearly, I seemed to miss a span or two and holy moley, what a mess. Fortuitously, I was able to find them all (the sgc's) and nuke 'em, but it made my head hurt.
I'll just use BD again, much as I hate adding that extra step. It's only 5 minutes and it'll save me this type of "SURPRISE!!!" in the future.
@Charleski: I'll try Atlantis again, I still have it here somewhere. Can I import .doc and export clean-ish html??
@Cap: good idea. I'll try that on the one I have in here for edits.
@Valloric: Still love Sigil.
@Ducky: Hola, comrade!! I feel your pain. Really.