My quick answer is MAP is a performance measurement. The area under the precision-recall graph. A face detection implementation will set a threshold and therefore, it is just a point in the graph. So this is two different types of measurements. I can have the same MAP value but two different curves.