2-state CDF fit is systematically wrong, but the 3-state fit seems OK. So maybe there is only a problem with the way the 2-state fitting is done.