Cryosparc has done a great job to help me achieve a high-resolution model. However, the corrected curve from NU-refinement looks quite wired compared to many published smooth curves. My understanding is that this could be the result of the heterogenous resolution of the map inside the mask, is this correct? In the final manuscript, which curve should I present?
Generally, you should report both the “Tight” and “Corrected” curves. The corrected curve always has a sharp dip at the resolution where noise-substitution was applied (in your case, around 7A). Some reviewers even look for the dip to be sure that the reported curve is a corrected curve. As long as the tight and corrected curves agree at higher resolutions (especially near to the nominal resolution of 3.6A in your case), the tight curve is valid and the corrected curve yields evidence that the masking did not cause any resolution artefacts.