Oh, I noticed a problem with the code you added to remove the extra <p> in Baen books, I tried to work out why it was happening, and in looking at your source I think I worked it out.
In your BAEN processing stuff, it looks like you have code that strips out the <a id="pXXX"> lines right before you try to strip out the whole <p> including them. I assume the second one never works since the first one has already modified the lines.