I wouldn’t worry too much about the discrepency you’ve observed, hker89.
While lazyness, mistakes or outright dishonesty can come into play, there are plenty of much more benign reasons for differences between a manufacturer claim and an independent test.
In this case, I think you’ve probably already hit on the likely source of most of the discrepancy. If you check ThorFire’s description, they don’t specify the chemistry of the AA cell tested. I’d guess, and robo819s retest suggests that they were probably testing with Alkalines. Not surprising, since a NiMH cell can deliver a lot more power (watts) than an Alkaline.
Another factor is that the numbers may not actually mean the same thing. I don’t know about robo819, but a lot of people test with a topped off cell. It’s the easiest way to make sure different tests are being done at similar state of charge. A downside is that it doesn’t give a very clear sense of whats typical. I can’t say for sure, but from comparing manufacturer claims to peoples tests of lumens at a given current and forward voltage, its clear to me that the number some report is more like the output when the cell is at its nominal voltage (3.6-3.8v). This is in many ways a much more useful number, but it can be harder to test for.
Yes I do test with the cells topped off always so I get the most consistent numbers and all cells are at their peak.
The run time testing is where I show what the cell is doing as it progresses, (as noted in charts in OP) as far as output.
I do not stop the testing once I turn the light on however to do any checking on the voltage or anything , as I always just turn the light on in the highest mode and then take measurements at the step down (sometimes at 30 seconds depending on light) and then usually at 15 minutes and 30 minutes and every 30 minutes after that until the Low Voltage warning , or cut off, and then check the voltage at that time.
Swap the tailswitch for a lighted one, ditto with a nylon washer and an opaque tailcap, and presto ……
Works fine out of the box, you don’t need to add a bleeder resistor.