Several years ago I ran some battery comparison tests mostly on alkaline batteries. I ran the tests becasue there were so many battery types at the B&M stores that it was confusing to know which manufacturer and battery type was the biggest bang for the buck. I have a RS DVM that has a RS-232 interface and can plug it into my computer and log anything the meter can measure. For the D batteries, I used a 2 ohm fixed power resistor which gives you 3/4 amp on a fresh battery and gradually drops from there. The tests were continuous until the battery reached 9.5 V. I tested Duracell CopperTop, Energizer Max, Rayovac, and Radio Shack batteries. Only 1 battery of each brand was tested, not a good sample I admit. I just wanted to get a general idea of what was going on. I looked at all the data from a several different angles: which one ran the longest, which one had the hightest battery voltage the longest, which one had the best runtime/cost ratio, and so on. Here's a quick summary of what I found:
Radio Shack: Ran the longest (21 hours), beat the next longest runner by 2+ hours. Most expenisve of the 4. Came in last for 0.8V to 0.5V.
Duracell Copper Top: Had the highest voltage output for the longest time. Longest run time from 0.8V to 0.5V.
Rayovac: Biggest bang for the buck cost wise, came in 2nd on runtime. 2nd longest runtime from 0.8V to 0.5V.
Energizer Max: All around poor performer, but had the 2nd biggest bang for the buck. 3rd longest runtime from 0.8V to 0.5V.
The reason for the 0.8V to 0.5V info was because some LED dropins that run on 2 to 9 volts will still be running if 4-6 batteries are used. Very important for very very long power outages. Battery leakage becomes a big concern at this low of a voltage. Use a cheap $2 flashlight with a NiteIze drop in and if the batteries leak, get a new flashlight. The drop in, which is the expensive part, will be protected and still useable.
As you can see, it's a mixed bag between batterys. Pick what parameter(s) are most important to you and use that brand of battery. YMMV. I didn't run any tests at lower currents since the runtime would be so long. Something like 8 days for each battery.