Probably, you have a dimer, may be homodimer which is numbered from 1. Maybe you simulating a dimer? That's why the RMSF graph counts the residues precisely as they are in the topology, and XmGrace basically performs an unusual thing where it attempts to link the different data sets, resulting in that weird pattern. If you have numerous chains, it is often a good idea to examine each one independently so that the data is in distinct files that can be displayed more easily.