What is weird is that they report that 3600lm while they measured 19,5A. Now is there a difference between lumen ‘out the front’ and purely a LED test, but in @koef3 his test of the SBT90.2 it did around 5000lm at the 19,5A mark. Sure, there might be some losses when actually used in a flashlight, but I can’t think of an explanation for such a big difference.
The 4600 lumens you estimate should be achievable (at least for a FET driven SBT90.2, but also with a buck driver, it should be able to at least pass 4000lm). The estimated throw could indeed be around 1800 meters, given that the Convoy 3X21D is pretty close in reflector size and lumen output and also reaches those numbers (more or less).
If it would really max out at 3600lm like 1lumen stated, then there would be something in the flashlight that is (seriously) holding that SBT90.2 back.