So by the time you reached that sentence in the article you must have gone over the part where this is a result co-authored by a geophysicist, and that the paper has gone through peer review of at least three other geophysicists (whom all have been contacted to hear their opinion on the work, and they're very positive about it), and that the geology community in general seems excited to see what practical uses this model will have in practice.

But sure, the fact that this is a model that starts with making the least amount of assumptions[0] completely invalidates it as geology and makes it pure maths. Because as we all know, no useful model in physics was ever built up from the ground like that, only introducing complex caveats as required to make sense of empirical data as necessary.

[0] the simplest way that one can cut a shape is with a straight line (2D) or flat plane (3D). In this case no new concave shapes can be created unless a concave angle was a part of the initial shape. Ergo, if the starting shape is convex and we cut away parts randomly, then all shapes produced this way will be convex. If the initial shape has concave parts, then on average most shapes derived from it will be convex. This is even more true if we apply this "cutting away" recursively to the shapes that were cut away. This makes the assumption of shapes being (mostly) convex quite reasonable, unless there is a specific reason to expect concavity, like specific crystalline structures perhaps. Also, for the record, I just reasoned this out on the spot, it's not exactly difficult.

Spherical cows in a vaccuum