[phenixbb] Rigid body refinement at low resolution

Hi all, I recently solved a complex structure at 8 Å. There are 3 molecules per asymmetric unit and each complex contains 8 domains. I was trying to do rigid body refinement in phenix.refine. However, all the rigid body refinement gave me worse Rfree, even at ultra-low resolution, and phenix.refine reversed the result to the original model. Anyway to make this rigid body refinement more stable? Is there a dampen factor in phenix.refine so that the program only moves a domain at a fraction of what it wants to, e.g. 1/2, 1/3… ? Thanks. Best, —Jianghai