Christof
You may be interested in a method I've used that seems to work very well. I have checked my Quatos calibrations and it has been very accurate. On machines like my Phantom I would trust this method much more than all the guessing I had to do with the mass distribution of the plastic fuselage. I think if you found the yaw moment and used about 80 percent of that for roll and pitch you would be in the ball park for a typical frame. Quatos is very tolerant and I add a 300 gram gimbal and gopro and extra battery's without any noticeable change. Adding the mass close to the center of the craft causes little change.
I use the equations from this video simplified to I = (.248 * m * T * T * b * b) / L
.248 comes from dividing the force of gravity 9.807 by the square of pi * 4
m = the mass of the copter in KG
T = the period of oscillation in seconds
b = 1/2 the distance between the strings in meters
L = length of the supporting string in meters
With the strings I'm using of about 2.5 meters long the period is about 2 seconds. It's easy to do yaw as it pretty much hangs level from the booms. All measurement are between strings and not motors. The string does not have to be at the very end but symmetrical about the center of the copter and the same distance apart at the top and bottom. It only needs to swing 15 to 20 degrees, if you do to aggressive a swing it starts to move around and you will enter the fact that it's lifting the copter slightly.