Will <blockquote class="calibre_25"><font size="5"> uniquely isolate all (and only) the chapters headings? You really need to find something that drills right down to the actual chapter heading text (in this case "CHAPTER 1"). You seem to be stopping just shy of that.
You may be beyond what the "Wizard" XPath expression builder can create automatically for you. Check out that XPath tutorial link that can be found on the Wizard dialog. It will help if you need to craft more complex expressions by hand.
Perhaps something like:
Code:
//h:blockquote/h:font[re:test(., 'chapter', 'i')]
Which will match all font tags inside of blockquote tags whose contents contain the word "chapter".
EDIT: also check some of the HTML in the various conversion stages (sub-folders) inside that debug folder if you can't get that expression to work. I'm not entirely sure at which stage of the conversion that XPath expression search will take place. And you may need to check the "Force use of auto-generated Table of Contents" checkbox, also.