Spin measurement retrodiction revisited