No luck so far - I downloaded a public domain book (plain text) from Proj. Gutenberg, converted to .doc and then a filtered web page to approximate my normal post-OCR procedure, brought it into Calibre, stuck a random photo (.jpg) from my hard drive on it as a cover (couldn't d/l one, it wanted to search by ISBN I think), made sure that box was checked (it already was), converted to .epub, and copied it to my phone. It didn't have the problem.
I'll keep trying though, maybe I can figure out something.