Dear Phenix community members, I have a molecular replacement solution refined agains a twinned 2.7 angstrom data to Rfree as 33%, with TLS and twin-law and others as default. The map looks good and shows extra map for my flavin ligand. Then I manually rebuild the model, fixing some obvious off-map residues and put a flavin molecule into the model. After this, I used the *def file from previous run to do another round of refinement. This time the Rfree increased to 36%. I realized that it has something to do with the TLS, isotropic and aniostropic B factor. However, I don't know exactly what's the reason and what to do at this stage. Any comments or suggestions? Should I do TLS refinement at the last step? Thanks and Best Wishes for all, Hua