I had the same thing happen to one of my XHP50.2, but I also have two still working fine : one in a Thorfire C8s with a H1-A driver and a Jaxman M2 with a LD-3 driver driven @ 6A.
Roughly 30% of the input power leaves as light, but this doesn’t mean the reflector is heated significantly. The reflector might absorb 10% of the light, so all together maybe 3% of the power is heating the reflector.
Presume a worst case scenario. Most useful input voltage value to make an output measurement is, in my opinion, the smaller voltage at which it still can maintain regulation. With an XM-L2 at the output most 7135 linear based drivers will maintain regulation at least down to 3.7V, and if PWM frequency is too high the voltage figure for mid and low modes will be even higher. Of course this is not based in 7135 regulators, but I said that to put this into perspective. That driver may not be able to stay regulated within an usual loaded li-ion voltage range (3 to 4V). This is how I believe we should teach these fellow testers how should they proceed. Thus, to answer your question:
Current is measured at 4.6V because the driver is a PoS and cannot stay in regulation below that voltage. Done.
Not exactly that, glide 1. That thing sort of pissed me off a little bit, I find it misleading. For a linear driver like that, I'd prefer to see the figures for ≈3.7V (or the minimum input voltage at which it keeps regulation -specifying the driven emitter-) and maybe somewhat below to see how it evolves.
This is not to say that 4.6V figure is not useful, it may be a useful figure for those aiming to use 4S Ni-MH input at the driver. Even in such a case, I'd complain because I always want to see worst case scenario figures, which for 4S Ni-MH would be below 4V.
Accepting those figures would be like accepting Ben Johnson's Olympic 100 Meterfinal record in Seoul, 1988. Sort of.