My runtime test of common cheap lights

I just purchased the last two reflectored Dorcy 1xAAA lights from my area Walmarts ($6 each). I had also recently purchased an Eddie Bauer branded Coast light from Target ($12). This one is called the "Mini Tac Torch" and uses 4xAG13 button cells and has a forward clickie. It is rated on their output system as having 24.7 mW, which is the highest of their button-cell lights by about double.
Since it's often said that button-cell lights have short and undependable runtimes, I thought I'd compare this one. Plus I wanted to check one of the new Dorcy's since it's known to have a decent runtime. I used the battery or batteries that were supplied with each light.
At the start, the Coast light was about twice as bright as the Dorcy. I think people are familiar enough with that light to know what I'm talking about as far as brightness is concerned.
At about 1.5 hours, the Coast had dimmed to the level of the Dorcy. The Dorcy had not dimmed appreciably. This trend continued until the 4 hour mark when the Dorcy quickly started to dim. At 4.5 hours the Coast, which had leveled off at a dim but useable level was once again about twice as bright as the Dorcy.
At 5 hours the Coast was still producing enough light to walk a wooded path, although you wouldn't have any leeway on it getting much dimmer and remain safe. The Dorcy was definitely too dim at this point to safely walk a wooded path although you could creep along if it was an emergency. It would have sufficed if your needs were lower, like the "Stuck in an Elevator" scenario.
Since it was time to go to sleep and the Dorcy looked like it was about to expire, I turned off the lights.
I've just restarted them tonight and both had rejuvenated considerably. The Dorcy quickly dimmed some but is still after 15 minutes giving enough light to walk a wooded path. The Coast button-cell light is about 1.5 times brighter.
I'll let you know the final results after they both totally give out.
The Coast packaging claims up to 100 hours of light, which seem utterly out of the question, but we'll see how far it can go. It has surprised me so far. I'm not sure what if any circuitry it contains.
EDIT: The Dorcy made it one additional hour during the second session before becoming too dim to safely walk a wooded path.

I continued running the lights last night with the following results.
As noted the Dorcy dimmed again after an additional hour for a total of 6 hours of usefulness with a break to recover at 5 hours. I continued to run it until 8 total hours to see if it would ever go out. It didn't but was making only a very small amount of light by that time. The battery had a no-load voltage of .68V when I removed it and measured.
The Coast button-cell light dimmed some after about 2 additional hours (7 hours total), but was still making enough light to see a wooded path. There was no margin for it getting any dimmer at that point though. I did continue to run it and it only dimmed slightly more. At the 9 total hour mark it was still putting out useful light but you wouldn't want to be forced to navigate at this dim level.
This has been an interesting test for me because the button-cell light performed so well for its limited battery capacity.
At least I now know how long I could depend on this light, even though it does dim a fair amount by the 1.5 hour mark. I was going to continue the test tonight but I've gotten home too late. I'll do it tomorrow night and let you know how long it keeps going.

CPFer Roy did a runtime graph of the old Dorcy 1AAA which more or less matches your findings. Replacing the LED in the old model Dorcy's with a modern high output 5 mm (nichia, smjled, etc.) will more or less double the output (I measured it..) without affecting run time. That why the old model Dorcy is one of my favorite lights (laid away a lifetime supply already...