Fix a rather nasty regression from r150690: LHS != RHS does not imply LHS->stripPoint...