Where are you coming up with that. No disrespect, but that mAh of the battery is not what you measure to see if it produces 750 Lumnes. The way to measure lumens is to first look at the specs for your bulb. In this case XM-L so here it is
http://www.cree.com/~/media/Files/Cree/LED%20Components%20and%20Modules/XLamp/Data%20and%20Binning/XLampXML.pdf
What does this say? It says that it can produce "1000 lumens with 100 lumens per watt efficacy."
What is the watts in your flashlight? To figure it out, do the following and use the formula below.
Formulas to use:
V= Voltage I=Current (Amps) R=Resistance (Ohms) P=Power (Watts)
Ohm's Law says V = I R
P = I V
How to measure Watts.
1. Measure Voltage first. Put the two leads on the "COM" connection of the tester and "Volts" or "V" connection. Turn your dial to "DCV" and put the wires on the two sides of your battery. If your flashlight takes two in line, line them up and measure both batteries. So two 3 volts should measure 6 Volts.
2. Measure Amps. Put the batteries in your flashlight. Do NOT put the cap on. Now REMOVE the wire from your tester marked with "Volts" and place it on the connection marked with "A" or "Amps". Turn the dial to your Amps setting. Now place the other ends of the wires on the negative side of the exposed battery and the other wire on the place where your cap screws on. The flashlight should turn on. You should be able to read the Amps. Now take the Amps measure from the flashlight (
note this has NOTHING to do with the mAh of the battery) and multiply that by the MEASURED voltage of your battery or batteries. Remember if two batteries in-line, then double the voltage. Now this number is Watts or POWER (
P = I V). This is the watts in an ideal world, but your flashlight not being made out of a good conductor (Gold) will loose watts, so you must deduct 20% of the wattage. Now multiply that number by 100 since the Cree manufacturer (pdf) states that you get "100 lumens per watt".
This approach is not a good way to estimate LED output.
The XM-L datasheet provides various graphs that are far more useful.
If you want 750 (emitter) lumens from a T6 flux bin XM-L, you will need a relative luminous flux of 750/280 = 2.7. The datasheet graph of relative luminous flux vs forward current tells us that you need about 2.3A drive. At that drive current, Vf looks to be about 3.25V, again based on the datasheet (see the graph of If vs Vf).
For a buck driver regulated light that uses 2xRCR123 to power the XM-L at 2.3A, let's generously assume 90% driver efficiency. That gives
Vbatt * Ibatt * 90% = Vf * If = 3.25V * 2.3A ~7.5W
Thus, Vbatt * Ibatt = 7.5W/90% = 8.3W
That means your power source (2xRCR123) has to be able to deliver about 8.3W of power to the driver, so that the driver can deliver 2.3A drive current to the XM-L to get 750 emitter lumens.
To figure out what 8.3W means in terms of Ibatt, a simple way is to assume at t=0 that Vbatt = 8.4V. Then at t=0+, the voltage rapidly sags to some lower voltage (let's say 8.0V), and finally holds at some quasi-steady-state voltage for most of the run time (let's say 7.4V).
Thus, Ibatt can start out at slightly less than 1A, increase to slightly more than 1A, and hold at about 1.1A for most of the run time. Let's just call it 1A.
1A is close to the limit for the current draw from a small RCR123 cell. If the driver efficiency is less than 90% (which is most likely the case), then the RCRs will have to deliver more than 1A. Stressing those little cells like that will sacrifice cycle life. I'd probably use IMR16340 cells.
If you want to power your G2 XM-L drop-in using 1x17670, then you'll probably use a drop-in based on a 7135-based linear regulator driver. The 7135 outputs a nominal 350mA drive per chip. If you get a driver that parallels 7 chips together, you get 2.45A drive, which is close to the estimated 2.3A needed to get 750 emitter lumens.
Using a 7135 driver means that your 1x17670 will have a current draw (Ibatt) of 2.45A nominal, as long as it delivers enough voltage to run the driver in full regulation (basically, Vbatt has to be at least 3.25V + 0.12V = 3.37V, and realistically more than that). 2.45A draw from a 17670 should not be a problem. Run time in regulation is probably going to be fairly short, however. Total run time will be longer.