I've heard that the technique is to raise the lower end of modulation range, causing the broadcast to sound louder even though the upper end is within limits.