Hi Pavel,
In upcoming version of PHENIX there will be more detailed manual for phenix.refine. Until recently phenix.refine was not using H atoms by default. We changed this behavior very recently. In newer version, if you have an input PDB file with H atoms (generated for example, with phenix.reduce), then phenix.refine will use these hydrogens by default. You don't need to do anything special for this.
OK. So I did the right thing by generating these hydrogens. If phenix does not use hydrogens at all e.g. generated riding hydrogens on the fly like refmac, it seems to me like the most likely cause of difference in clash scores. I don't like to see those hydrogens in the coordinate file though as they are mocking my resolution.
Analogously to wxc_scale in coordinates refinement, to play with B-factors restraints use wxu_scale.
I tried reducing it but I didn't see major improvements in Rfree-R difference.
Please look another post on phenixbb (Subject: "Re: [phenixbb] R- factor problem !!!") regarding how tricky and error prone could be to jump between different programs.
I haven't found it that tricky. I have to grep away the ANISOU records and perhaps reset B-factors before TLS refinement. In my limited experience however the TLS refinent seems more stable in phenix. With refmac I tend to get negative Us quite easily. Especially when changing the model. I understand that for comparing statistics or R-factors, but if you refine the structure, it is just a bunch of coordinates. Scaling is another issue. I've had cases with ice rings and xtriage tells me there is no problem. Refinement in phenix goes okish, but to get decent results with refmac, I had to cut away those resolution shells. I guess I'm going off topic now. Time to do more test runs, ~L~