You have the gist of it. The load control DynoJets are hybrids. It's an eddy current brake used to load the engine for writing maps but isn't capable of steady state torque readings. It has a load cell that needs to be calibrated via an arm and weights. So it's physically capable of it. Not sure why they didn't go all the way.
It really doesn't matter (within reason) how accurate the numbers are as long as they're consistent. I don't consider it done until I get 3 consecutive runs with the same result. It's use is as a comparator so you can see the results of changes. An increase/decrease will give evidence of the path your on being right..................or not.
The DynoJet uses RPM,gearing and the acceleration rate from A to B to calculate the numbers. The RPM comes from an inductive sensor attached to a plug wire. Gearing is calculated by RPM vs. drum speed rate that is read via Hall Effect sensors. On set up it asks how often spark occurs,i.e. 180/360/720 degrees to help determine gearing. The software is calibrated for each machine via the drum serial number that allows them to determine the exact weight and diameter via records from manufacturing. The DynoJet dyno is better than it's detractor would have you believe. It's a solid tool that in the right hands allows the best to be extracted from and engine. The operator is the standard that makes it right. A good operator takes care of his machine and knows how to use it.
That being said notice the Factory Pro EC997 is my favorite. It's faster,lighter and easier to get consistent results from. It uses a far more reliable 5 gas analyzer vs. the DynoJet O2 sensor which has serious shortcomings. However those standards come with a price. A very steep price.