Thanks for the praise, and for taking the time to write such a detailed post.
I understand the need of a HTML-centric generation mode. However, this is not the aim of this program : it's called Calibre2Opds for a reason
For the moment, the OPDS catalogs are generated (using the defined options) and then they are processed by a XSLT processor to generate the HTML catalogs.
It may change : I was thinking of rewriting the program in a plugin-driven toolchain, where the output would be generated by a specific plugin, which means that everything would be possible, but also that generating two (or more) different formats would take twice (or more) the time. But even this is not on the shortlist of upcoming features.
I'll keep it in mind, though.
What I can do is try and give you all the features you want, that do not penalize the OPDS catalog generation. They are, not exhaustively, allowing for bigger thumbnails, letting the user choose a XSLT template (or a CSS stylesheet) to customize the generated HTML output, adding the generation date, and finally adding an "All books" catalog not split by letter.
If you have other requests, please state them, and I'll tell if they are reasonable or not