Very nice analysis! A lot of the specs raise serious eyebrows, and this test of the SBT90.2 gives everyone enough information to see what is off.
(1) 65A turbo. The test shows the SBT90.2 completely maxxed out by 30A (under optimal thermal conditions!), so I’m curious to see how the hell they can push double that through the emitter, and what happens to the emitter when that is done.
(2) 6800lm. The bare emitter doesn’t even touch 6000lm; factoring in a typical optical loss of 15% in a reflector light, 5000lm is all one can reasonably hope for.
(3) 2126m of throw. This is not even theoretically feasible. Using the 3x21D as a reference, 1800-2000m is a much more probable estimate, depending on whether you want turn-on vs 30s intensity, and whether you use tgreviews or 1lumen reviews as a reference. A simple upper bound for intensity (lux) at 1m of a Lambertian emitter is given by Lr^2/A, where L is the output, r the radius of the optic, and A the LES of the emitter. Taking L=5000 lm, r=45mm (still overestimating reflector size), A=9.6mm^2 gives an upper bound of 1.05M lux at 1m, or 2054m ANSI throw.
All in all, I don’t think this light offers any meaningful improvement over the 3x21D, especially given the price tag, and the difference in quality of customer service. The 3x21700 configuration is more than enough to comfortably drive a single SBT90.2 (no meaningful gains above 20A), and can be used in other lights, as well as being more easily replaceable. In the event of battery failure, I shudder to consider what a 46950 can do.