I shouldn't complain too much. 150 lines of Java automatically take filtered HTML output from Word, apply custom styles, remove DIV, SPAN and FONT tags and break the mess into chapters in under a second. Saves me hours.

It takes me longer now to clean up the mess Calibre makes of the ePub conversion. I really need to finish my MOBI to ePub generator so I can eliminate Calibre from my process altogether.