In my experience, M-3 greatly overestimates the conductivity, and thus the distance to field strength contours, of AM stations. The F(50,50) computation used for FM is based on a truncated model, and assumes relatively flat terrain, and also overestimates the distances to field strength contours. For various reasons, most people, actually a great majority of people worldwide, live next to the water-oceans, lakes, and rivers. There is usually a relatively flat area next to the water that extends in several miles perpendicular but not always. Island areas from volcanic sources can rise up very rapidly. Those coastal and shoreline areas may be close to the predicted field strength. Once you get into the hills and mountains, signal attenuation becomes much greater, both a shadowing and a signal bending effect. If those areas are dry and sandy or rocky, the conductivity is even less.
M-3 Maps were constructed from Class I groundwave measurements and Class III DA proofs from the earliest AM DAs. Those also tended to be in coastal or shoreline areas where the conductivity approximates M-3. Many AM stations have moved out further from the coastal and shoreline areas, far from the coastal plains, and the conductivity is less there. Nondirectional Class IVs in small inland towns did not normally measure conductivity, and hence no conductivity data. The M-3 Map has not been revised since around 1960. That is probably why so many new AM DA arrays have disappointing service. Measuring the site conductivity before building a DA has become a lost art. Land cost, zoning, FCC and FAA regulations, etc. have become overwhelming considerations at the expense of technical advantage.