Evidence that HTML as a master format cannot be satisfying, or at least not universally so. Seems we already knew that, but it's still sad to hear that your toils didn't have the intended outcome.
I have happy results when I process that HTML on my own machines using my own tool chains, in part because I can, and do, easily check my output from those tool chains. In fact it has gotten to the point where if I want to read some book from PG by some other submitter, I find that I can *easily* create a more satisfying read experience locally: 1) Download the PG HTML file(s) 2) Edit the CSS to fix the half dozen formatting "errors" that almost all submissions from DP fall prey to. 3) Compile locally using the latest version of Kindlegen, DIRECTLY submitting the HTML files -- no epubmaker required! 4) Read and enjoy a book that *isn't* scrambled. If I can do this with 5 or 10 minutes effort, then PG could too. Its just a question of whether PG keeps defending the status quo, even though its clearly not working [1], or when or whether PG is willing to acknowledge "you know, the approach we have been trying to take simply isn't working, we need to rethink our approach." [1] How do "we" know that the current approach is not working? Simple 1) Open a PG book in a web browser. 2) Open the same book on an epub device 3) open the same book on a kindle device. Are the results similar, and similarly satisfying? Nope. Are others able to accomplish this same task using PG books as feedstock. Yes. Conclusion: PG is doing something wrong.