Yeah li-ions can go lower and the built in protection is oftern lower then 3 volt. The old train of throught was 2.7v per cell min for lipos now its 3.2 volt min per cell and some new ESC cut out higher. The batteries seam to last longer and the internal resistance stays lower for longer
I used to test my li-ons to 3 volt as habbit and most my torches cut off around 3 to 6 volt but the guys on the forum spoke up so now i go below 3 volt to try and help every one.
Another thing to take into accound while working out the run time of a certain torch with a certain battety is, the higher you discharge the battery the less mAh the battery will give.
So if you discharge a 2400 mAh 18650 @ 3 amps it will not gove the full mAh its will give around 75% of the claimed mAh (2400) i was not going to bring this up as i felt the math was close enough.
There is so many factors in this, DDM's, batteries, the charger, the flash light its self, the user and where the batteties are brought from and there protection PCB's.
A hobby charger reall helps with this this as you will know the battery cut offs and excatly what the battery will do under the same draw and the bettety the battery the more it can give.
That is not correct, but depends on the actual chemistry in the cell and what cutoff voltage you use.
Here is one example of a battery where the capacity difference is very low between different currents:
HKJ, I was thinking more along the lines of trustfire cells i used a 2400 mAh 18650 trustfire flame as an example and under rated my results. The reason why i used 18650 trustfire flames 2400 mAh as an example as JOE who asked the question said he was using 3000 mAh 18650 trustfire flames that we know only test between 100 to 200 mAh more, so the results would be closer and they are the closest batteries i have to what JOE was talking about. The results would be different with sanyo, panisonic, IMR or 26650 cells.
2400 mAh trustfire flame: 3 amp discharge test from fully charged to 3 volt.
I thought this was more relivant. There is so many factors in this, DDM's, batteries, the charger, the flash light its self as each flash light, even the same Model may vary, the user and where the batteties are brought from and there protection PCB's and the flash lights driver efficiency.
The general info on here will give a basic aprox run times for flash lights based on the current draw on the tail (switch). I have done a few 0.5 amp, 1 amp and 3 amp discharge test,s on the 18650 2400 mAh trustfire flames since they test really close to the 3000 mAh flames you should get a ruff idea with a bit of math and a current draw test on your torch.
Run On Full Until empty and Check how long it took to kill t h e m....... 
It depends very much on the batteries, I did test some good 3000mAh and some not so good 2400mAh:
Also try placing a line at 3 volt, you will see the difference at different currents is greater, especially on the 2400 battery.
You might also see a few batteries where the mAh is slightly greater at 3-5A current.
I see whay you mean, your 2400 flames did not test to well at all, but the 2600 (3000 flames) seam about right, nice snd grouped. Where did you get your 2400 flames from DX ?
As i found the manafont trustfire flames tested higher in the mAh department then DX and with the 3000 & 2400 mAh flames brought from manafont there was less then 200 mAh between them as the best 3000 mAh flame tested just over 2600 mAh and the best 2400 was not far behind.
I brought a few trustfire flame 3000's for my old but since i was not getting more then 200 mAh more i whent back to buying the 2400's as to me it was not worth the extra buck for a few seconds more run time.
Any way your graph is much better then my data table as is has all the diffrent current draws on it so say if the flash lights low voltage kicks in at 3 volt like you say run a line across it and there you go.
My data table just shows if you gun a trustfire flame 2400 in a light with a 3 amp drop in you will get 42 mins run time to 3 volt.
So JOE not knowing what torch your testing if it has a low votage warning or if it is a muti cell light, you can use HKJ's graph or you can use your imax b6 set it to discharge at the same current draw as your light with the same cut off as the torch or 2.75 volt for the battery and discharge the battery, when done the charger will tell you how many minuets i took and how much mAh the battery gave.
If your unsure of the cut off voltage most off my singe cell lights have a 3 volt warning or none the light might dim, my muti cell lights is around 6 volt for 2 x 18650's. So 3 volt or 2.75 is a good starting point just make sure the cell is full charged and rested before discharge.
If you notice you discharge a battery to 3 volt then once the charger stops and the volts rise up to 3.3 volt thats normal they do that and the longer they rest generaly the higher they climb back up.
The 2400 was from DX and the 3000 was from Manafont.
In the reviews I have also charts that directly shows the runtime, here is one of them:
But that is only correct for lights with a constant current consumption, in most lights the current will vary with the voltage.
The time in mins has to be one of the most important graphs and why i use data tabes.
For these data I prefer charts, it is much easier with all the runtimes in the same chart and you can interpolate very fast. I.e. a light that uses 2.5A and works down to 3 volt, it only takes seconds to estimate runtime from the chart.
I do have many megabytes of tables for each battery, but I seldom look in them.
This beats all the other more complicated ways being mentioned. If you need to know the runtime of a specific cell and torch, just test it!
Ps this min voltage cutoff, do any of you guys adjust your cutoff voltage relative to current draw? Manufacturers of certain batteries specify it is safe to go to a lower cutoff under shorter tests or higher under longer tests. Not sure if this applies to 18650s but i imagine it will without checking.
Nice graph! There has been a lot of helpful and insightful information in this thread from everyone.
HKJ, it appears that the TrustFire 3000mah batteries PCB kicks in at 2.8v, correct? Also you wouldn't happen to have a formula for the discahrge time would you? Just to figure runtime on lights don't pull exactly 1.0 ,2.0 , 3.0, etc.
I have some Flames 14500's from MF they go into low voltage protection at 3v, or that is the measured voltage of the cell when i pull it out of the light. I guess that voltage could be a recovery voltage since there is no load at the moment?
No, my test stops at 2.8 volt, the protection is at a lower value.
That is easy: capacity/current, you select the mAh from the curve closest to the actual current draw (From the capacity chart) and then divide with the actual current draw.
For the TrustFire 3000 you would just use 2500mAh and with 2.5A draw, the time would be (mAh must be converted to Ah) 2.5Ah/2.5A -> 1h.
You can also just look at the "discharge time" curve and guess, that will probably be just as precise. (The variation in current draw with voltage spoils any precise calculations).
This method is my original method but it just doesn't seem very accurate. Or is 2500mah really the "usable" capacity of these batteries? I mean I know they can hold that. I just assumed that the "usable" value (or until ~3v) would be a least 100mah less than overall capacity.
That is the reason I say you must check the discharge curve to get the real capacity and us a curve close your actual current level.
I am confused about flashlights that have 2x and 3x batteries.
For example a 2x18650 with a tail-cap reading of 1Amp. Using 2500mah batteries. What would the estimated battery life be? 2.5hrs or 5hrs?
For batteries in series (This is the most common configuration) it would be 2.5h.
Ok, so lets say my STL-V6 gave me a tail-cap reading of 1Amp. That would mean the emitter is getting 2amps and the estimated battery life would be 2.5hrs.
What about a 3x18650 (DRY battery configuration) with a tail-cap reading of 1Amp using 2500mah batteries?
That is also a series connection, both batteries and leds, i.e. the current is 1A in all batteries and all leds. Again the estimated runtime would be around 2.5h