Darned article! It's the bad penny that keeps coming up. I (and several others) thought it was right myself and I posted about it until I did the math and found out that the frequency of banding we're getting on the Solidoodle (18tpi) can't be related to step error. If microstepping had an effect it would not be spaced matching the thread pitch but at an intermediate one based on the layer height you were using.
I calculated that microstepping ribs would be spaced out every 0.5mm when printing .1mm layers, and another spacing when printing at .3, and none at all when printing at a multiple of .07055mm. And yet on the Solidoodle everyone is seeing ribbing at a constant frequency of 1.411mm no matter what layer height they print at.
Thus, if this sort of microstepping-based ribbing is occurring, it's far too small an effect to notice on the Solidoodle, and is being far eclipsed by the wobble-based ribbing.
Instead, all the tangible success I've seen is using combinations of one or more of the following: flexible coupler, soft motor mount, thin rod, fine-threaded rod, anti-backlash device, and software compensation.
Using a fine-threaded rod is my current favorite; since you're changing the thread pitch, it's the only method that's guaranteed 100% to eliminate the ugly, coarse, 18tpi banding. Even if you still had the worst wobble/backlash in the world, bands at .5mm are hardly noticeable compared to 1.411mm, and going to a fine thread and thin rod automatically reduces backlash and wobble.