I would not trust the output numbers tested for large lights like the GT110. The issue is that integrating spheres are designed (1) with small opening to measure small lights, and (2) to measure diffuse rather than highly concentrated sources.
In comparison, the Lumintop X9L with the same LED claims 6500lm but achieves only 4812, and this number is more realistic, in line with the test of the LED here. If you perform a quadratic extrapolation of the test curve, the LED would have trouble breaking 5500 lumens even when highly overdriven. Add in optical losses and non-ideal power supply and cooling setup, and no SBT90.2 light could realistically break much past 5000.
The battery capacity difference is significant, but I doubt it’s noticeable in real use unless you plan to run turbo continuously for hours at a time. It’s like recharging my phone once every 2 days versus once every 4 days: the difference is significant, but really does not matter in practice.
I’m still most worried about the single high-capacity battery: 38000mAh at 3.7V is 140Wh; in comparison, a typical electric bike battery is only 400Wh–this cell contains as much energy as 1/3 of a e-bike battery pack. And many buildings won’t even allow these bike batteries to be charged inside, for good reason. Unlike a single cell where a single defect could make the whole cell fail, multiple cells with the same combined capacity is less risky as it is much less probable for more than 1 cell to fail simultaneously, so even in the event of failure the amount of energy released is only a tiny fraction of the combined capacity. I don’t trust Lumintop’s battery and charging circuit to be safer than that of electric bikes.