I have searched the archives for the converse of the "how big is too big for loop modeling?" issue but haven't come across anything. I have modeled a protein that has two 1-residue and one 2-residue insertions relative to the templates. If the insertions were larger, I know that I could use loop modeling to further refine the model, but would much be gained by loop refinement of such a small region (would 1 residue even count as a loop?)? Would extending the "loop" by several residues on either side of the insertion be appropriate in this case, even if it means allowing residues with corresponding positions in the template structure to move?