Hi Ethan,
So far as I know, there is still no provision in phenix for reading in existing TLS parameters. It always wants to regenerate them for itself.
You are right, given suggested optimal selection for TLS groups phenix.refine is capcabale to do the rest as good as it can. There is simply no need to input any values of TLS values to phenix.refine since they will be determined based on actual reflection data and not user supplied B-factors), so just atom selections defining TLS groups is all phenix.refine needs. In one the next versions PHENIX will automatically determine the optimal TLS groups partitions as part of refinement and this is what will be used it in refinement.
Cheers, Pavel.
