Early on a reporter had a trip to see the Band team and I do remember reading them say that the GPS was used to calibrate a users stride.
That said, all the website info says about the Band enabling advanced analytics, and so far that has to be seen... I don't know if anyone has evidence of a band calibrating a users stride, e.g. run before enabling GPS and then run with GPS and then run without GPS again and see if the distance is better on the third run (all on a clean band/account.)
My first run on a treadmill had my pace at around 90 seconds per mile slower than the treadmill said (9:10 vs. 7:30 minutes/mile). I then ran outside with the GPS on and it matched my running watch (Garmin 610) within .01 of a mile for the 2 miles before the battery died on my Band (left the stupid charger at work on accident). Today, I hit the treadmill again. It was closer (8:49 vs. 8:00) but still not where I'd like it.
At the end of my battery life using GPS, I was running a bit slower pace, 8:33, when it died. So, it may have used this as my baseline for calibration (not sure if it averages out a full run or takes the last .25 mile). Generally, when I'm running faster or slower, I'm really only adjusting the length of my stride, not the number of steps per minute (unless I'm at a full out sprint) so it might have assumed my 8:33 stride length in the calibration. I'm going to continue playing with this but I suspect it will rarely be very accurate if you alter your pace at all. I had the same challenges with the Garmin foot pod which essentially does the same thing.
The other thing I noticed on the treadmill is when I took the time to look at the band to get my pace, by stopping my arm movement enough to be able to read it, I noticed that my pace slowed down. The treadmill didn't change so clearly the less swing in my arm made a difference.