[incrudible]

The problem is that your suggestion is strictly worse and unnecessarily complicated for the case discussed in the article. If you want to downsample 2x, 4x, etc, then that's just one level in the MIP hierarchy, no need to compute the rest. The point of the article however is to explain how one level in that MIP chain can be computed in the first place.

If you want to downsample 3x or fractional, then interpolating between two MIP levels is gonna be worse quality than directly sampling the original image.

Perspective (the use case for anisotropic filtering) isn't discussed in the article, but even then, the best quality will come from something like an EWA filter, not from anisotropic filtering which is designed for speed, not quality.