If you continue to read, the article points out the architecture differences of the various CPUs and GPUs and how routines are farmed off to one or the other as needed. This is where the problems start: shoving a routine off to a processor that is not designed for it is like going to the plumber for a broken elbow - he may be able to fix it but it will be slow, painful and ugly. The reason we have a separate processor for graphic routines is because it's made to do it both faster and better. This is not rocket science.
The larger problem is that consoles are a fixed architecture and coding for their constantly aging components requires more and more optimization to continue to up the wow factor. Consequently the newer console games will actually perform worse when ported if the code is not optimized. This would explain why games aren't even ported to begin with - the code is too proprietary.
So sloppy porting is simply lazy coding, as most suspected, but it explains why a faster CPU may make the graphics in a Ported game run smoother instead of a faster Video Card. It also explains things like the PC version of Assassins Creed and GTA IV using lower res textures than the 360; why rewrite code when you can just compensate cheaply, even if it is only (an est.) 10% of the codebase.