That’s a proven way to go dating back to P7’s but some had issues because of the lower Vf of newer LEDs. It might be less of an issue with the higher Vf XML-2 but in any case people using this circuit have to pay close attention to battery voltage since the driver won’t read this correctly any more. It also works better with single mode drivers(no IC).
Diver in the above pictured Seoul P7 setup still has to burn excessive voltage for all three leds. But I guess if it worked with two additional batteries and P7 leds it will work with one additional battery and XM-L led.
I think you’re probably better off going the master-slave approach, personally 
Batteries will be in 2s/2p (8.4V) and will connected to headlamp with 1m (3 feet) of cable. A master/slave configuration would require batteries in parallel and twice the current through the cable which would also mean 4x the loss.
Using a 0.75mm2 cable (the most I can get through cable gland) that would be 0,4W vs 1,6W (3A or 6A current)
P.S.
Thank you all for info
go for an Lflex then, tailor made for this set up.
The 2280mA KD 7135 V2 driver with 2 batteries in series and 2 XML in series worked on high for ten minutes. It got hot but not so much that I couldn’t hold it between my fingertips.
Wait really? That thing should have poofed... Are you sure you didn't wire them all in parallel?
I wired them as pictured in my first post.
Ok, well I'm gonna set up something to test that now (I have a pile of pre-XR-E LEDs). Be back in a bit.
Damn, I'm out of 7135 based drivers, does anyone remember how I solder it up with a bare 7135 chip?
AMC 7135 has three legs:
ground
output ground Vin
Its hard to tell from the pictures, but I'm with scaru, that shouldnt work. 8.4V on that driver is too much and I dont see an LED in the path of the cells.
On second thought: are you sure the driver is working? Have you tried modes? Maybe you bypassed the driver.^^
That is actually very likely, I just tested it and it seems to have instantly killed the chip.
Damn it, it does work! :D I had a bad connection somewhere in my set up, but I got it working and it is regulating it perfectly! I'll get a picture up in a few minutes. After that I will leave it on til it burns. :evil:
I think it might be working because the current from Vdd to ground is very small so the power loss is in that part of the chip is low. I am surprised that the IC didn’t blow and when you try it on low it may still poof. The way that mod was done in the poor man’s multilux , one of the LEDs was placed between the batteries and the driver to drop the input voltage to the IC. The version I did last summer replaced the 7135 closest to the led+ with a 7805 (2981 for 2cells) voltage regulator and rewired the the traces to give 5V to the IC. 
The 7135 in the blue box is moved or removed and the VR chip goes in that location.
Instead of removing the diode it can be used as the jumper from input + to the VR chip. This provides the whole unit with reverse polarity protection.
All modes worked and I also tried it on 6 AA (Energizer lithium) but I only left it on for a minute or so.
Any update on how it went?
There were some who you said “you can’t do that” to me so I won’t say that to you but just suggest you carry a spare which you probably do anyway. I did a fair amount of testing with that mod and all I can say is that whenever I exceeded 6V to the driver or chips something would fail, usually with a puff of smoke and a fzzzt. Just skimmed that thread and the highest I went with complete success was 3 LEDs and 9 nimhs. Like you I was looking to eliminate the power loss from the high currents required by parallel mods and use the stock switches on my lights.
No, I had to recharge some batteries first. But it is currently running and has been running for the past 2 minutes.
12 minutes in and still alive. 