Now writeClusterGraph uses rmsd for similarity measure instead of GDT (maxcluster was failing almost always with GDT)