Dear All, By phenix.real_space_refine, I have fitted the a PDB file to an EM map, with Ramachandran favored around 90% and acceptable level of Ramachandran outlier (1.5%) and rotamer outlier (1.2%). Besides visual checking indicating the fitting of the main chain and sid chains to the dnsity map is acceptable. I hope the refine results can be improved by phenix reciprocal space refine. By phenix.map_to_structure_factors I convert my mrc file to mtz file, with the above mentioned phenix.real_space_refine refined pdb amd mtz files as the input files for the graphical phenix refine, in which I never select real space refine, and I always check xyz coordinates and Group B-factors (sometimes I check xyz coordinates and individual B-factors ). Fot this kind of phenix reciprocal space refine, I find the molprobity anslysis results have improved a little, but I find both the contouring level of 2FOFCWT map and the FOFCWT map is more than thousands of thousands high (and it seems the contouring level of FOFCWT is much higher), and visual checking indicated both the main chain and side chains hardly fit with the density map. But I really find there were publications which did phenix.real_space_refine first, and then process the reciprocal space refine, which get good results as introduced in the publications. Will you please tell me why my phenix reciprocal refine after phenix.real_space_refine did not work? Best regards. Smith