Thanks for the feedback for creating Mobipocket ebooks from HTML. I gather there are two areas of concern, the underlying .html and .opf coding. I will look at your changes and see what I can use/compromise/discard when I re-make a .imp ebook from it. It will better help find that "common ground".
HTML (even filtered HTML) from Word can be a "nightmare". I usually try running the Word HTML through TidyGUI and select the Configuration setting to handle that the source file is from Word 2000+. I then search and replace large spans of my default formatted paragraphs to strip the in-line style <p style="">. It does a decent job, but not as good as initially starting from .html.
One trick I've used in the past with great sucess is when starting with a webpage or similar .html, copy the displayed text and paste it into a blank email created by MS Outlook Express 6. Then click the Source tab at the bottom of your email message and Select All and Copy that HTML 4.0 code into a new text file opened by your favorite text editor. Then save that file as your starting HTML base for the ebook. It avoids using Word, but requires some quick search and replaces to get rid of http:// references to images and requires you to manually copy images to the source directory. Just food for thought.
Oh, by the way, any chance of showing me what you would/could change to make my above .lrf/.epub version better suited to your own preferences. Can you post a .lrf to see the results? with source?