Keroberos and Hitch, I didn't do the initial conversion - have no idea how it was done - although it appears to have originated from a scanned physical document. There is not much I can do about the initial state of the documents, I just want to be able to "clean" them and have them stay clean - which seems to be happening now, with DiapDealer's suggestion.
huebi, I had the same problem, but it doesn't seem to be happening now that I've turned off HtmlTidy.
Rob Lister, I no longer have examples - and I'm not eager to reproduce by turning HtmlTidy back on - but one situation I saw was <p><span>...</span></p> where the ... represents plain text. I would delete, save and then have the <span>...</span> reappear when I reopened the file. Maybe I am overlooking some steps on my part that contributed to my problems, but I'm very satistfied with the solution.