During 3D refinement jobs (heterogenous, homogenous, NU, local), I have to use a larger window inner radius (0.95 instead of 0.85) to avoid masking of my map at one end. I noticed that in heterogenous refinement only (but not in homogenous, NU, and local ref.) the larger window inner radius is in effect only at the final iteration. I.e. when looking at the real space slices for intermediate iterations, my map is still masked at one end.
On a related note, can someone explain how cryosparc handles image/map recentering?