I was curious to see how great an error Don's method produces. It does seem to get you pretty close on most shots. Below are three plots of the errors in impact direction at three different CB-OB distances, measured center to center. I included the 8" separation out of curiosity, but he explicitly recommends that special adjustments be made whenever the balls are less than 15" apart.
The horizontal axis is the intended impact angle, with divisions of 5 degrees. This is the desired OB direction with respect to the CB-OB line of centers. The vertical axis is the error in degrees, and each division is 1 degree. The curves come to an abrupt end on the right, where a greater than 90-degree cut angle would be required to increase the impact angle. (The corresponding cut angles, not shown, are always larger than the impact angles.)
Most of the errors are undercuts until you get to extreme cut angles (where the red curve crosses the horizontal axis). Overall, it looks pretty accurate, but anyone using this might want to do some tweaking, especially in the 25-55 degree range, which I think he also suggests.
8" CB-OB Separation:
17.5" CB-OB Separation:
36" CB-OB Separation:
The geometry would be perfect if you moved the plane in which you "double the distance" forward or backward, so that it intersects (includes) the center of the ghostball. But then of course, you might just as well aim at it! Obviously, that wasn't his intention.
Thanks as always for making this available Dr. Dave.