1. Reflector vs lens:
Properly focused reflector will capture more light than a properly focused lens. (Well, it’s actually independent from whether they are focused properly or not).
According to Enderman’s calculators deep reflector can capture about 60% of light. The rest is spill.
With a lens even 40% is a huge stretch. With a F# if 1 you’re capturing under 17%. The rest is lost.
2. XHP35 HI vs LEP
XHP35 HI gets 170 cd/mm² at the max. LEP does 400. You need to increase head diameter by 50% for the LED to compete.
3. Single vs. multi-emitter
In theory, using multiple emitters is often the best answer for the longest throw / size. But I define size as overall light volume, which boils down to head volume.
Why? As you noted, if you compare a single-emitter light to a multi-emitter and keep emitters constant, light emitting surface gets bigger which may cause thermal and electrical hardness.
Also, multi-emitter lights also have lower effective reflector area for the given head diameter.
The latter can be somehow countered by using configurations that maximize head usage. Like 7-up. Then you get a fair use of frontal area. Then you scale the diameter up to have the same effective area as with a single-emitter. You end up with smaller head volume yet the same throw.
The former can be countered in theory by using smaller emitters. LEPs are small. The smallest LED that we know to be good is a little over 1 mm². Quite big already, so in practice it’s not so great as in theory….because it adds up to 100 W near-peak with a 7-up. Needs a hefty host.
Also, focusing multi-emitter light might be harder. How precisely can we solder emitter to MCPCB? I have a hunch that the error margin may be comparable to light emission surface diameter….