I've been using GRT for several years with for an AR15 in 223.
After entering the velocities for 7-8 shots [i.e., doing what CharlieNC said], GRT's OBT calculator is usually within +/- .4 grain of the actual nodes; sometimes within .2 grain and, occasionally, right on.
Note:
Right on' for me is getting groups 1/2 MOA or less.
With my AR [and over 20 or so different barrels], I've found uncalibrated GRT to overestimate the muzzle velocity by 50fps or so. IIRC, Sierra told me that AR's tend to run slower than bolt actions by ~ 50fps.
When I go to develop a new load, because I don't know how close the predicated powder amount is to the actual node, I load up 5 loads .2 gr apart - 2 loads above the GRT predicted charge and 2 below and 1 at the predicted amount.
Since this covers 1 full grain spread, and nodes are 'never' more than 1 gr apart [at least for the size groups I need], you could just pick a charge and do a 5 step ladder.
The reason I use GRT is, after calibrating it with the velocities from the 7-8 shots, to see where the max charge would be and arrange my 5 step ladder so I'm safely within the max.