On the fidelity issue, we have created the OO stylesheet to look like a classic book where possible (except for the coloured lines), but in this version we were unsure about how to put leading lines above and below a block extract for example, and then adding a first para style for the non-indented paras so they look good in the OO file. That will be more XSL work, but make the using/learning curve steeper. We will look at that in the future.
Interesting point about the XHTML source code, and I may blog about that a bit further, the difference between the Structure-Styling and other HTML environments, is that it always looks the same for the same structure, so in some respect you don't really need to see it (if you believe this!). I am not sure that this is the place to get to technical on this matter, but I might put an extra "chapter" into the online tutorial. So assuming the XSL's and slicers and dicers are working nicely, the XHTML elements and class statements are totally predictable and I can just crack open the style sheet, or make a whole series of custom style sheets for a range of looks and feels.
In one development version we did have an XHTML exporter, but thought that got too complicated. It output the book as a single, fully processed XHTML file - the one that is used internally before being split apart to create the final package.
Having said that, I think we were trying to address the make a good looking eBook fast with a bit more styling for someone who would prefer never to see the source. (are there any of those on this forum?)
From the believe it or not department; the last point is our shame!
We are a Linux development house working primarily in Python and mainly do Web services applications. Some silly little interface issue stopped the deb, but its coming.