Made much more progress today:

• Stripped down the XHTML document which has my lookup tables in it.
• Added most of the new sequences we have to account for, based on comparison with the printout yesterday, to that document.
• Rewrote the XSLT that processes that document to produce the conversion XSLT so that it does all its escaping correctly; the results can now be directly copy/pasted into the conversion XSLT. You can imagine the double- and triple-escaping required when you're trying to recognize and convert characters requiring to be escaped into new sequences also requiring to be escaped.
• Reformatted the output of the above, so that each operation appears on a separate line; this makes debugging much more practical.
• Clarified the difference between the Greek ypogegrammeni diacritic, and the combining dot below which is used to indicate doubt about the identity of a glyph in a transcription. Both need to be accounted for.

I'll probably get to the point where all the new sequences I know about are handled by the end of the afternoon, and I might have time to run a test. I'm also going to try converting the font over to Gentium, which is what the publisher prefers. Hallelujah -- a publisher that wants the same thing as we want.