Progress report:
Okay, I am still performing data analysis. There was a fuel leak from the IP return at the banjo fitting (rather vigorous at that) at the end of the first run which pretty much ruined the physical measurement of fuel via graduated cylinder readings. With the amount of vibration present while the engine ran on the dyno however, I'm not sure how accurate that data would have been anyway. It's probably a blessing in disguise that it was lost--as it may have mislead us anyway.
Now the good news is Brian (JetPuf) logged a ton of vag-com data: fuel temp, coolant temp, IAT, fuel consumption, RPM's, speed--and all are time-stamped by the computer. Since the dyno ensured constant loading, I am hopeful that a valid
relative comparison can be drawn, once I normalize the data. Normalization puts the data on an equal footing, relative to the thermodynamic importances of engine temperature, fuel temperature, and IAT.
I'm going to get help from some of our engineers like GoFaster, Ernie, and TdiMeister on how to best normalize the data.
With non-normalized raw data no meaningful comparisons can be drawn--it's utterly inconclusive. At this point all the non-normalized data proves is that engine efficiency is STRONGLY correlated to coolant temperatures, and less strongly correlated to fuel and IAT temperatures. The heat-soaking treatment proved much harder in reality, simply because the largest graduated cylinder I had was a one-liter one. Most of the fuel was consumed in bringing engine coolant temperature back up during the first run! Also it took longer than expected to swap lines and not have leaks (lesson learned: screw-type hose clamps pinch and generally suck, and so do zip-ties for small lines

)
I'll get this dyno test data ready as soon as I can--will be done in a few more days.
Also, Brian (JetPuf) and I have thought out a more robust road test using 1 gallon containers and measuring the fuel consumed after the runs are complete (in the lab) and using vag-com to record data. We think it will be confirmatory to whatever we find out from this data set.
What might be a good idea is if a few of you who have vag-com would perform the test on the road identically to what Brian proposes--he's going to post on that soon.
Edit: new additions.