I searched for the answer to that when I first started using Dirac. I don't recall whether there was any official documentation on this subject, but after reading several discussions, I have been since pointing the mic to the ceiling and using the 90deg cal file. Dirac does three consecutive L-R-L sweeps and doesn't give me any time to reposition the mic to the next speaker, so I don't really understand why I should point the mic 0dge at any speakers, let alone between speakers. I understand the argument of positioning the mic parallelled with the floor/desk to capture more reflections, but that means all speakers are off-axis to the mic, and I would have to compensate for that when adjusting the target.
The speakers I own have anechoic and Spinorama data available, so I can compare Dirac's sweep result with them. At least for my setup, pointing to the ceiling at 90deg yielded a flat high-frequency response that is similar to that from the anechoic or Spinorama data, and Dirac will suggest a target that overlaps with the sweep results pretty well. If you can find anechoic or Spinorama data for your speakers, you can compare them to your Dirac sweep using different mic positions and then decide which one is for you.