I’ve been using using these XPath expressions in Calibre to get the table of contents from the epub generated from Push to Kindle with great success.
Level 1 TOC (XPath expression):
//*[name()='h1' or name()='h2']
Level 2 TOC (XPath expression):
//*[name()='h2' or name()='h3']
Level 3 TOC (XPath expression):
//h:h3
I believe this would be a great addition to Push to Kindle!