Over lunch with my supervisors (aka advisers) managed to sum up everything I'd failed to communicate in our friday meeting in a single sentence.

If I prove that markup by PPM is a strict subset of markup by HMM and write code to do the translation then comparing them becomes much easier, since we can run the two systems using the same engine on the same problem with the same settings.

If only I'd had the presence of mind to put it like that on friday.