It's often worth considering an extreme example. Suppose, for instance, that we put virtually all of the mass of the moon into four extremely dense "mascons" located near the surface and at the vertices of a square (whose center would, therefore, necessarily coincide with the center of the moon.) Now consider the net gravitational force on a satellite as it orbits at relatively low altitude in the plane of that square. You should be able to convince yourself that the orbit would consist of four relatively low curvature sections when the satellite is far from any of the mascons (and, therefore, subject to a relatively low gravitational force) and four relatively high curvature "corners" as it passes each mascon (and is, therefore, subject to a relatively large gravitational force.) If you are visualizing the resulting orbit correctly, you should see that the satellite might graze the surface in between mascons and reach maximum altitude as it passes each one. The orbital shape would reveal the location of the mascons.
Repeating the gedankenexperiment for different orbital parameters or for different distributions of such high density mascons should make the general principle clear.