What do the occupancy fields look like in the PDB for those loops? They’re probably 0 occupancy. The PDB reader is probably throwing the residues out as invalid (since they weren’t in the electron density from which the structure is made).
You could run them through fixbb instead of idealize (use NATRO for the resfile) – if the same issue occurs it means this is occuring at PDB read in time, not within idealize mode.