Hmm.. I'm asking what the actual text headings are. But if they're in bold you can probably write an xpath for it. I had a similar book, and I used this xpath (under the structure detection preferences):
Code:
//*[((name()='b' or name()='h2') and re:test(., '\w{2,}', 'i')) or @class = 'chapter']
That will look for any bolded text with two or more word characters and mark it as a chapter. (most of the alternate tests aren't needed, I just modified the defaults slightly)
Alternatively you can post your book over on the bug tracker - bugs.calibre-ebook.com. I've been working on some new Chapter markup/splitting routines for pdf conversion, I can use your book as a test case, since it doesn't seem to work with the current functionality.