TOC in pdf export only uses pagetitles

The Table of contents in the PDF is generated based on pagetitles, not based on the headings in the document. This means that the document structure does not necessarily match the table of contents.

For example, lets say the document structure is as follows, each separate wiki page is marked with a star

|- h1. Introduction*
|- h1. Basic use*
  |- h2. First command*
  |- h2. Second command*
|- h1. Advanced use*
  |- h2. Extra command
|- h1. Final words*

Note how the Advanced use wiki page contains the subheading Extra command, which is not a separate wikipage.

The table of contents will be rendered completely omitting the h2. Extra command.

|- h1. Introduction*
|- h1. Basic use*
  |- h2. First command*
  |- h2. Second command*
|- h1. Advanced use*
|- h1. Final words*

What I would expect is that the table of contents for the pdf would be generated based on the hX, since those describe the document's structure. Also it makes the table of contents completely useless for long wiki pages.

In our case, we are using CSS rules to add numbers to the headings (which wasn't too easy to figure out how to do). However since the toc isn't generated from the document strucuture, while the CSS rule for numbering is, they get out of sync when some headings aren't in the TOC.