If it helps, that curve seems pretty typical. IIRC the FL-1 standard measures brightness at the 30 second mark. The only thing I don't understand about the curve is that it apparently spikes *up* around the 40 second mark, where I would expect it to be heading down.
I'm not sure, but I think the 40 second mark may be the point at which the light ramps
down - 10 seconds seems like a reasonable buffer for gaming the ANSI system. The ramp-up would be shorter in that case, I think I remember hearing it's about ten seconds of ramp-up when the light turns on?
There may be flashlights out there which can sustain full output until the batteries run down but I don't think I have ever seen one, except for the ones with quite low output. A modern high-output flashlight can run 5 or 6 amps through the LED and a good chunk of that power is going to end up as heat. A bigger/heavier flashlight body like the Maglite will be able to soak up more heat in the body before the driver has to start throttling, but even that is only postponing the inevitable unless you have screaming fans or a chilled water hookup.
One trend I have been seeing more of recently is to label the high setting as "turbo" (which feeds more power into the LED than the heatsinking can sustain) then have a "high" setting pegged at around the point that can be sustained.
Still not sure if I like that though - it adds another mode, and one can make a good argument that 99% of users would rather have peak power for the first minute or two even if it was unsustainable than to artificially limit the output to a sustainable-under-all-conditions (eg high ambient temperature) level.
I don't necessarily care about super high output, I've had my priorities rejiggered twice already now and lumen rating has gone down both times (I don't reckon I'd realistically need any more than maybe 400 lumens). Reading into it, it seems candella and, to a lesser extent, CRI are way more impactful on a light's performance. Not sure how exactly candella relates to lumens or how a light is tuned for one vs. the other, though.
A note: I mentioned wanting a 3-cell light before, but having thought about it, reconsidered, and thought about it more, I think a 4-cell (given either D cells or 26650s) would be more suitable as a talking stick. Maybe less portable, but that's not a Maglike's role - if you're carrying it around for a long time, you don't put it in your pocket, you get a holster for it.
Re: turbo, I actually kinda like the idea of a "GIVE 'ER ALL SHE'S GOT" mode for emergencies when you really wish you brought a bigger light, but didn't, though I agree that a simple UI is better - the simplicity of my Maglite's mechanical switch is hard to beat (especially since there's zero parasitic drain). If it's there it should perhaps have its own toggle/activation button, "behind" the mechanical switch so to speak. Or otherwise just not get in the way and leech precious electrons. I suppose I dislike misleading output figures more than I dislike turbo mode fine print, ultimately.
re: upgrading an older Maglite, my impression is that the Malkoff drop-ins (which are unfortunately out of production AFAIK) were the gold standard since they used a screw thread to force the LED's heatsink into good contact with the flashlight body.
I have an 800 lumen Firefly upgrade and it works well (adding low/med/high modes makes the light a lot more useful) but I suspect you would see a similar degree of throttling to what the graph above shows or worse... the new Maglites have a pressed-in metal thermal path from LED heatsink to flashlight body.
You're right about the oldschool Malkoff dropin being out of production - the one with the securing screw. They make adapters instead that let you drop in the emitters that they use for their other lights. But they seem too heavy on lumens and not heavy enough on candella. Their claim of 4h +2h per extra cell on alkaline D cells seems somewhat promising, if true, but again unless anyone can say otherwise for sure, I don't think they can handle the extra voltage of 3 or 4 Li-ion cells, or could do so without going off script in performance. They could also have a bad case of ramping down/throttling (as you mention) either rapidly or over time, which renders all my speculation largely pointless.
Re: the Firefly, I was thinking about the less hot 235-325 lumen one mounting the Luminus SST 40, to clarify. That's the one that's just the emitter slug and uses the stock reflector. Same concerns as with the Malkoffs, basically. And they don't even list the candella, actually...
I made a note of that Convoy light for future reference when I decide I need a new one
Browsing the interwebs, I came across the Nitecore TM39, and the Lumintop BLF GT - performance wise, those are two lights I can definitely get behind, but they
still don't manage to be quite what I want! They're both inappropriately shaped for reverse-grip use, have complicated electronic UIs, probably weigh too much (hard to tell without having one in my hands, but 1.3-1.8 kg
sans battery seems like it'd start feeling about 10x heavier real quick), and a little on short side for use in contact-based communication (or fire poker, or general-use "ten foot pole", etc)
It's funny, sourcing a good flashlight that ticks all the boxes I need is proving
disproportionately more challenging than I thought - finding suitable battery cells was way easier, in fact I pretty much wrapped that up from start to finish since I came back last Wednesday, even not spending much time on it. I found a couple American outlets that'll ship to Canada, and I've got my pick between protected or unprotected Keeppowers, up to 5500 MaH, and unprotected Vapcells, up to 6200 mAh, all flat-top 26650s. Easy enough to turn into button cells with some small magnets if need be. Now I just need a flashlight to put them in!
