3. I see that phenix.refine has been used in the internal runs, I have used phenix.refine earlier.
Autobuild uses phenix.refine: it iterates between model building and refinement with phenix.refine. Fore more details see this recent paper that describes it in details: Acta Cryst. (2008). D64, 61-69 Iterative model building, structure refinement and density modification with the PHENIX AutoBuild wizard T. C. Terwilliger, R. W. Grosse-Kunstleve, P. V. Afonine, N. W. Moriarty, P. H. Zwart, L.-W. Hung, R. J. Read and P. D. Adams
But after playing with the side chains in Coot, when I tried to use the modified pdb file and resolve.mtz, it complained about the good old Free_R flag.
It seems to be a bug in PHENIX. Dale Tronrud reported a similar problem (look mails with subject "Using the Same Test Set in AutoBuild and Phenix.Refine"). If phenix.refine stops with "MD5" error message, a temporary fix is: remove the corresponding line from your PDB file and use your original reflection file in refinement. Let me know if you have any questions. We are looking into this problem and hopefully it will be fixed once Tom is back. Cheers, Pavel.