AMAZING job, Jon! Thank you!
Wow! This would mean the TS10 ate almost the whole battery, while the D3AA still had ~28% left!
(interpolating from the table @HKJ posted here)
This is IMO a huge advantage for the D3AA! Goes to show how important/useful an efficient driver is (and I hope @Wurkkos and @Wurkkos_Terry are reading this!)
TS10 stepped down to sublumen after 34.5 hours, the battery reached 2.90V and triggered LVP.
The D3AA battery is at 3.54V… and holding near 5 lumens still
Again, the D3AA is simply amazing here. On an emergency situation where one needs continuous light and doesn’t have many batteries nor the possibility of recharging them, this could be literally the difference between life and death.
When the TS10 and D3AA step down from long runtimes at 5 lumens, it happens suddenly. Essentially giving the operator no warning that the battery is getting low.
This is really bad, and while that behavior doesn’t surprise me with the D3AA (with a regulated driver, that’s not a bug but a feature),
I expected differently from the TS10, ie a constant dimming all the way, following the battery voltage. This for me removes the only advantage of an unregulated driver… 
AFAICS, all of these would require the operator to periodically turn off the light for the checking, which would hinder operations.
I wonder whether we could/should implement a 4th method: the light, when turned on, would constantly monitor the battery voltage and then start blinking the Aux LEDS (in case they are viisible, eg as in the switch in the D3AA) or the main LED (for the TS10, where they’re in the optics along the main LEDs and therefore would not be visible) when the voltage reaches some predefined value (eg, 3.35V as it would mean 10% remaining charge, interpolating from the same table I mentioned above). Of course, in case of the main LEDs, the blink would need to be sporadic (like once every 20 seconds) in order not to hinder the light’s usage too much. What do you think, is that a good idea?