I'm assuming here but most likely by taking the rifle/load out to long ranges (1000 yds or so) on paper and walking it in to the correct dope, then putting the atmospheric info along with the actual amount of elevation into a ballistics program to "true" the dope. This typically will back calculate the muzzle velocity and adjust it to what is happening in reality.
The calculation essentially compensates for error in either velocity or BC. Doesn't really matter which you want to consider as your error. But I've found that with certain brands of bullets I have less error which leads me to believe the published BC is where the errors are, more so than the chronograph, but I'm also aware enough to know its most likely some of each.
I'm currently running the newest version of the magneetospeed, so I can't really comment on other versions either but would like to run some side by side tests of mine vs an older generation unit and see what we get.