Mismodeling in gravitational-wave astronomy: The trouble with templates