Dear Cris,
Have you tried the various anti bump macros? There's an example in Alan's cime-decompose.inp which should be in one of your example folders and is also online at:
https://community.dur.ac.uk/john.evans/topas_workshop/tutorial_organicsolution.htm
Syntax is:
AI_Anti_Bump(S, S, 3, 1, 4)
and it's described in tecref along with other box_interaction possibilities. I think in practice over-use of these penallties can actually prevent you finding a solution and in many cases the data should pull you to sensible non-overlapping solutions. Others have far more practical experinence in this than I do.
The other possibility for the H's might be locating them by DFT post-solution if their positions are realy important?
John