Sig - yes controlled experiments are the way to go. A good test should have a control running in parallel if possible. There should be calibration runs before and after the experiments. First to check the instruments and later to show that nothing has changed with them during the testing.
For good IP transfer there should be several steps. One - have the " inventor" run the experiment while watched, two - have the experiment run by someone else under the direction of the "inventor" (without the "inventor" actually turning knobs) with the inventor's device, three - have the replicator make the system from "scratch" (nothing supplied or altered directly by the "inventor") and run it.
If you can't get to #3 then the tech was not transferred. Try again.