Your interpretation of how the Treatment Analyzer works is correct. It heuristically explores different allocations of a fixed budget. There is no optimization given a specific objective. I would suggest using an iterative approach where you try the treatment analyzer a few times in a row until you hone in on a treatment combination that works best. In the past I have found that even one iteration can be quite informative about the activities that have the highest return on investment.
Note that in the latest version of Path (4.2) this feature is called the Transition Target Analyzer. Let us know how it works for you!