I read this whole thread with interest as I am now doing tests for KFX. I can report a few things:
1. There is absolutely NO evidence that KFX is erasing or in any way ignoring the NBSP. I used it to bind ellipses and em dashes to the preceding text. I also used it to bind a traditional three-point ellipsis, all to see if KFX would break any of them up or force an ellipsis or em dash to the start of the next line. No matter the display font size, I could not force any to break. I have attached two screen shots. In the first, I have circled where the text would have stayed on the previous line had the text not been attached to an ellipsis. In the second, the last letter, d, has been pulled to the next line precisely because of an NBSP.
I also used decimals ("1.5") and 4-digit numbers with a comma ("3,000") to see if KFX would split any at the period or comma; it didn't.
2. In my test file I:
- embedded a font
- tested for the Monospace font
- used media queries for older mobi7 content
- created lines of various color, style and width
- created tables
- used hanging paragraphs
- added inline and stand-alone images, using both inline styling and CSS-controlled divisions
- created colored text boxes
- used selective colored text
- used Emphasis and Strong with and without font-styling (i.e., emphasis without Italic)
- created ordered and unordered lists, and mixed the two
All were done in HTML4 and CSS2 in an ePub2 in Sigil and converted in KFX. Everything displayed as intended. I cannot know if the Emphasis without Italic would read properly in TSS as of course the file cannot be tested outside previewer.
I then exported a mobi file from Previewer3 and sideloaded it into my older Kindle Fire. Everything displayed as expected.