How do I know? The measurements quoted above. In this torch, XP-G2 makes the most light at 350mA, the highest regulated level of current. Results may vary at other currents.
My quick measurement also agrees with djozz’s tests. The test shows more lumens from XP-G2 for the entire graph, compared to 219c, until the point where the current exceeds what XP-G2 can handle. This is defining efficiency only in terms of lumens per amp though, not lumens per Watt. Because lumens per amp is what matters on this type of driver.
XP-L HI isn’t on that graph. Normally I’d expect XP-L (domed) to have the highest lumens per amp, but perhaps the XP-L’s advantage doesn’t start until it gets to higher currents? Or maybe the HI’s lower bin eliminates the advantage? Let’s check the measurements… Going from djozz’s test data, I see…
- XP-G2 S4 @ 1A: 440 lm
- XP-L HD V6 @ 1A: 490 lm
- XP-L HI V3 @ 1A: 400 lm (estimated)
- Nichia 219c @ 1A: 390 lm
But at higher current, XP-L HI may be more efficient…
- XP-G2 S4 @ 5A: 1330 lm
- XP-L HD V6 @ 5A: 1820 lm
- XP-L HI V3 @ 5A: 1486 lm (estimated)
- Nichia 219c @ 5A: 1200 lm
The estimated HI values are the HD values / (1.07 3), since Cree specs each bin as 7% brighter, and HI is three bins lower.
It looks like the intersection point is at about 2.8 amps or ~980 lumens. … which I think works out to about 11 amps and 2850 OTF lumens in this light. So, in the D4, XP-G2 is likely the most efficient up to about 2850 lumens, and then the XP-L HI is the most efficient for higher levels.
XP-G2 has the highest Vf though, so it’ll have the most sag on a low battery. Nichia 219c should have the least sag.
For practical purposes, they’re all good choices… so pick whichever one has the tint you like. Tint is much more noticeable than the other emitter differences.