Min-P, Max Exaggeration: Critical Analysis of Min-P Sampling in Language Models