Hi,
In the “new” local refinement in 3.1, each part remains in its original location - not recentered in the middle of the box. This makes map combination and other downstream operations such as particle subtraction more convenient.
For map combination, aligning maps to a global reconstruction and combining by taking the maximum value at each voxel (vop maximum in UCSF Chimera) is the best method I have found to get a seamless combination. I would still encourage deposition of the orginal component maps as well, though.
Cheers
Oli