Shipping a bunch of different Makefiles isn't necessarily worse. Consider that in the intervening 20 years, all of the extant platforms have converged significantly, due in no small part to POSIX and SUS. Today, apart from major feature differences for which there's no simple solution, there are fewer incompatibilities. Papering them over requires some ifdef's, yes, but it's not that bad. Using separate Makefiles (1) allows automatic builds to work in the common case, and (2) simplifies fixing issues because the learning curve wrt the build for that project is shallow.
In the modern context, the value of autotools and similar is less, and the relative cost higher.
The bigger issue today is remaining "portable" between Linux and Windows. That causes many times more headaches than in even the most backwards of Makefile hacks.