I don't know the answer, but an educated guess.
Maybe dialog normalization?
I'm not sure that DVDSP sim mode respects it as I have some audio differences when it is used. Just to be sure, re-encode some assets without diag norm and see if that helps.
Tech-T Productions
http://www.technical-t.com
ps-None of my comments are meant to be gospel. Otherwise, I would write a book that provided all the cold, hard, fast answers that everyone assumes are out there. These are the things that I have found to work well with my shooting habits and workflow. Always try things out for yourself and deduce the best workflow for you.