On the validity of Martin's real-part formula