I purchased single samples of a few types of cells from IMR and have been experimenting as well. Here is what I see so far.
First, I removed a small amount of weight, perhaps 4.5 grams, from my second SV in the form of the battery holder (See my battery holder post) and the top plastic shroud (no clips, just 4 screws after you remove the bottom cover and the camera board). The prop guards would add about 10 grams if I ever used them so I figure I should have about 15 grams headroom without exceeding stock weight. This thing seems to fly pretty close to it's weight limit to begin with so every gram counts. I've also noticed with the heavier cells, it can't maintain hover below about 3.1 volts. This is not just because of the internal protection in the stock battery. I haven't yet checked the logs, but I suspect if you do you will see that the highest output motor in this condition is at almost 100%. I also set FS_BATT_VOLTAGE to 3.0 for my tests. Voltages were measured via wifi/mavlink telemetry. My test hover is performed outside and sometimes includes some wind so results do vary. I also do a short rapid climb every minute or so to get a feel for how much headroom we have on motor power and to make times more realistic than a static hover. The extra current draw from the climb is sometimes enough to drop the voltage/power enough to cause the end of the hover. Reported weight includes JST connector and wire. I solder directly to positive terminal and side of battery case (as quickly as I can to avoid heat spread). I suspect such abuse is frowned on but I have no way to weld it and I bought cells without tabs. My conclusions and measurements are below:
Stock battery, rated 1200 mAh: 27g, Based on testing I also find that about 9 minutes is maximum safe flight time, flys very well with fairly comfortable power:
Reduced weight SV flew for 10:12 before test hover could no longer be maintained. The voltage was about 3.0v immediately before landing but bounced back to about 3.4v just a few minutes after landing. I put 1123 mAh back into it afterwards. No significant swelling of the cell, but I suspect I've reduced it's working life. No cell should be expected to run this close to it's rated capacity for many cycles. I won't be repeating this test on a stock cell.
Sanyo NCR18650GA 3500mAh 10A: 48g, No significant gain in flight time over stock. Flew OK but feels sluggish and climbs a bit slowly:
Voltage drop due low current rating is limiting factor.
LG HD2 18650 2000mAh 25A Battery: 45g, Provides 12-13 minute flight times. Flew fairly well, perhaps a bit sluggish:
Hovered until 3.0v, 13:45 flight time required 1515 mAh charge. Second test hovered 12:53, required 1480 mAh. Capacity/voltage drop were limiting factors.
Sony VTC6 18650 3000mAh 15A Battery: 49g, Provides about 15 min flight. Feels sluggish, climbs slowly:
Hovered until 3.1v, 15:45 flight time, required 1770 mah. Current draw/voltage drop/weight were all limiting factors.
EFAN 20700 4300mAh Battery: 63g, Rated 15A. Could barely get off the ground and flew for only a few seconds:
Just too heavy and not enough current!
Sanyo NCR20650A 3100mAh 30A Battery: 53g, Flew for perhaps 3 minutes. Didn't bother recording the details:
It looks like 45g might be about the sweet spot. At this weight, we can lighten things so we are only about 3 grams over the weight the stock drone flys indoors with the prop guards without removing parts.
At this weight, it looks like we need a cell that can supply 8 amps all the way down to 3 volts. The power which can be supplied at 7-8 amps before we reach this voltage will determine the maximum flight time.
I also have a 30c 2000 mAh 2S lipo pack on order which should come in at about 45g after I separate the cells. (Why is it so difficult to find high current 1s packs?) I'm hopeful this may provide a reliable 15 minute flight with some headroom.
I may try some more Li-Ion cells as well if I can find some more on the light side with high enough current. I'm guessing they may be less prone to damage through high drain/low voltage.