I tend to get varying results with FoCal more than the original versions when it just came out. I think the more and more complex the program gets, the less and less reliable it is becoming.
I used to run lenses over and over and get the same MA value. Now I get a wide range no matter how I test... 25x FL, 50x FL, floodlighting, natural shade light, bright sun, large target, small target, different papers, etc... My 24-70 II for example at 24mm at first said 0... Then -5... Then -3... Then +1... And the results all look very good, graphically. But it just changes each time I run it.
35L was the same way, as is the 70-200 and whatever else I have tested with. I don't enjoy using it as much as I used to honestly. Btw, my 5D3 has been checked/calibrated as well as my lenses by Canon and I don't have any real world focusing issues so its not my equipment.
I have for sure noticed it has gotten less reliable, though. I know many will probably disagree but I remember the original program (when it was still an original beta for 5D2 only) would nail the same value over and over for me. And my real world shots proved the results. Now, not so much.