I tried to combine my overall map(A) and local refine map (B) in chimera using vop maximum command and got composite C. At the same time, I used Phenix combine focused map (map A & B, and a overall pdb as inputs), got composite D. However, I found the local refined region(B) is different in C and D. A obvious shift was observed(see picture below). which result I should trust? Has anyone seen this problem before？
you should not trust either result. The combined map is not necessarily a true representation of a structure that actually happens in nature. It is the result of moving and shifting 1 region of many particles until they come to a common shape, and doing the same on a different region, and then applying user input to stick them together. Composite maps are for illustration purposes only. The Phenix version is biased by the pdb model, and the Chimera version will stick the two together wherever they are. In chimera I would typically fit-map the subregion to the global refinement and vop maximum. There is no reason to build a model to either of these.
Does that mean, to build a overall model, I should fit local map in overall map in Chimera first , and then build overall model in coot according to these two aligned maps, but run Phenix real-space-refine using overall map and local map separately？
I would build and fit the model to the local map, and then take that to build the rest and refine in the consensus map. Though the local map is your best data for the masked region, you only have sufficient data to build the whole model in a consensus reconstruction and that’s that. You can also, and I would, refine a model specific to the local map in the local map and deposit the local map and the model consistent with that density.