It sounds to me like Google is doing pretty much what you would want it to do: keeping the html versions in the index, and the PDFs as supplemental. In this case, I don't think that the supplemental classification is bad, since the pages are actually different URLS.

I also don't think that this is going to cause any sort of penalty WRT duplicate content. However, if you are really concerned, you could block indexing of the PDFs with robots.txt.

Thanks for the reply. That was my feeling as well.