The dynamic factor model revisited: the identification problem remains