Hello, fellow grad student!
I assume you’re talking about the Davis & Baker paper.
As a general rule, with the constraints – you just have to sample to figure out what works. It’s likely to be a system-specific result; the more the system is flexible, or the xtal structure is poor, or there’s binding site motion, the weaker constraints you’ll want.
Your question is deep enough that you may just want to start poking around in the code to figure it out. It would be better if we had perfect documentation, but we don’t…
I’ve contacted the author to see if he has further insight.