With the Fenix PD35, as 2x CR123 will provide a higher voltage than 1x 18650, would that mean it'd run brighter on CR123s?
Review here:
http://www.candlepowerforums.com/vb/showthread.php?381867
Although it might not directly answer your question.
The reality is, no probably not. Or not be enough to really tell visually by eye.
What is more likely that on Turbo the higher voltage will offer longer sustained output, but overall much shorter runtimes.
To explain, the XM-L2 requires an input voltage of approx 3.x volts. A fully charged Li-ion 18650 has a resting voltage of 4.2 volts. As you apply load to a battery (amp draw), it will suffer voltage sag.
This means on a fresh battery it will provide Turbo mode fine, as the battery voltage drops, the voltage sag will mean the battery voltage may well drop below that required by the LED. So it's output will drop.
You can see this easily on this chart, note how the blue line drops quickly:
So even though the battery might have quite a lot of mah capacity left, it won't have the voltage to offer the Turbo output, only the lower modes.
This depends on the driver being used in the torch however (3 main types; boost, linear & buck). I don't know for sure, but I don't think the Fenix uses a buck/boost driver looking at the discharge plots.
I think it is just a buck driver of some kind with maybe a linear mode for single cell use.
By using 2xCR123a or even RCR's you start with a much higher voltage than you need. The driver in this instance will buck the voltage down to the level the LED needs. So even when you include voltage sag as the batteries are discharged, you'll generally stay above the 3.x required by the LED. So this isn't giving you a brighter light in PEAK lumen terms, but means it'll offer a brighter light for longer. See this discharge chart:
Note the blue line doesn't dip as sharply and is flatter.
And even more so on RCR's:
But the thing to note is, overall runtime including lower outputs, the multicell arrangements will generally last less time. e.g.
The 18650 starting on turbo manages over 240 mins before completely shutting off (follow the blue line on the first chart, note the step downs). But on RCR's it's more constant output on Turbo, but only manages 40's mins before shutting off. A whole 200 mins less runtime overall. The CR123a's are in between the rechargeable options. Also note the RCR's give the longest turbo runtime before stepping down and the flattest, but shortest overall runtime.
However there is a caveat here. Selfbuilt tends to use a massively outdated 2200mAh protected ICR 18650 to test with. This not only is low capacity for a modern 18650, but also low amp draw. You have two choices:
1. Run a high performance IMR 18650
2. Or a high capacity ICR 18650
For example, if you used something like a 2500mAh Samsung 25R 18650 IMR battery. It has a very high discharge rate and will suffer less voltage sag than the 18650 used in the test. This means this 18650 would offer a more stable discharge and more light for longer. Probably not to the same level as the higher voltage multicell options. But vastly better than the 2200 used in the test. Total runtime would still be much longer than the multicell options too.
If you went for the 3400, discharge performance isn't as good as an IMR, but is better than the 2200. So Turbo performance will improve still, just more marginally than with the IMR. But total runtimes, especially if you use lower modes more often will be massively longer, maybe half as much again.
I hope this has given you enough info to allow you to answer your own question.
Personally, unless I had specific need to use CR123a's or were given them for free, I see no reason not to use a good 18650. Much less hassle and cheaper in the longer run.