I appreciate the heroic efforts to measure total light flux as reliably as possible. However
only total lumens and OTF lux are insufficient to characterize the beam.
I think it’d be better to measure “intensity” (in the photometric definition), as a function of angle instead of only along the flashlight axis. If we had a table of candela X angle we’d have a much more complete picture of the beam and could answer questions like how big/intense is the hotspot, how much spill or corona there is, etc. This is unusual because it’s too much for the general public but for this community it’d be very useful and it seems that measuring lux reliably is a lot easier than total emission.
In principle we need measurements in the whole plane but almost all flashlights have a beam with cylindrical symmetry so readings along a single line are enough. It can be along a circle or a perperdicular to axis.
If we know the distance to the flashlight we can also easily integrate and determine total lumens emitted in the measured angular range, and going far enough even total lumens. This way we wouldn’t even need an integrating sphere and if we have such measurements we can compare with the angular result.
These readings take some time, so it’s better to do them at low/mid levels to avoid output drop. Since the beam profile won’t change with intensity level, doing 1 or 2 readings at max is enough to scale the complete graph.