I just got the software a week or so ago. Here are some observations.
[list type=decimal]
[*]Read the manual. Maybe not every page, but there is some good information.
[*]The "fully automatic" adjustment is quite tedious on the 5D III and any other camera without support for AFMA adjustments by software.
[*]TurboCal is much faster than "automatic" - I can't say whether the results are consistently as good.
[*]For the big lenses you are going to need a big indoor space, or the ability to work outdoors when lighting is consistent. I had trouble with tree shadows moving across the target before I could complete a single calibration. It might also work to have the camera outdoors and the target in a controlled indoor space with an open door or window.
[*]My results were mostly similar to calibrations I had attempted by hand. I'm taking that as a good thing - what the software says agrees approximately with my eye. Probably the software is more accurate.
[*]Speaking of accuracy (or really, consistency) I did some "semi-automatic" testing and it was surprising how much the results varied from one shot to the next under identical conditions. There was often much more difference between consecutive shots at a single AFMA value than between two shots with an AFMA setting several units higher or lower. This suggests that taking several shots and choosing the best (when you have the luxury of doing so) will make more difference than a perfect AFMA setting. Of course what I saw with one lens/camera combination may not apply elsewhere. By the way, this test was not done with the 70-200mm lens mentioned below.
[*]The autofocus on my trusty 70-200mm 2.8L IS lens died while using the software! I suppose it was ready to die after 7 years of use and who knows how many hours with the shutter button half pressed. Still, it was a shock. Both autofocus and manual focus are now unreliable.
[/list]
If anyone else has tips about how big to print the target for various lenses and target distances, I'd like to hear them. Also, target setup with TurboCal seemed different than with the other methods, and I'd like to understand that better.