Originally Posted by JimmXinu
It's really weird, IMO, how calibre treats HTML books. While it renames the files for all the other times, it keeps whatever the original file name was for HTML. Inside the zip file, that is, that it forces them to be in--but doesn't unzip again for viewing. It does follow links and images--but then it flattens the structure and changes the HTML to match.
If I remember correctly, Calibre converts all non-epub files to epub for viewing. Perhaps the behavior you're seeing is because of that?