Reconsideration of saturated D-optimal design