Test #3 settings are the settings we use at studio , where we have strong lights , but at my house where lights are lower , images were darkish and so i dont know if this test has any meaning at all.
We're not interested in the images, that's why the shutter speed is fixed (I guess you edited it to 1/40th for test 3) Otherwise, if you shoot in dim lighting, the camera might choose say, 1/2 sec exposure which would throw off the results.
The
content of the image may affect the result, because the file size can change a lot depending how much detail there is.
I made an xls with formatted cells , because the produced csv file is a bit messy 
In general, I'd rather have the original data, but no need to re-post it for this run.
interval = 1000msec (i know that camera cant get this time , so it will shoot as fast as possible.)
It looks like this was actually 0, but that has the same effect. The code where it says "for PTP testing" is only used if you didn't run the script from the script menu.
What i noticed is that apparently different sd cards , dont have any impact on shooting speeds if i read the results correctly.
This does seem to be true in this case, the total variation is only ~200ms and most of that does not seem to depend on the card.
Some observations:
The CHDK benchmark "write (64k)" speeds are quite close to the nominal speed class. The class 10 card does much better with larger writes (raw and mem)
The first class 4 L results are slightly odd: If the image is ~5 MB, it should take > 1s at 4 MB/s. But the ready_time - raw_time is only 660ms. So maybe there is buffering going on, with the actual write finishing in parallel to the next shot.
You are right that test 3 is a bit faster, even though the exposure time is ~25 ms rather than ~0.8ms. You can see the longer shutter shutter in raw_time (910 vs ~880-890 in most other runs). I expect this difference is due to the ISO, because ISO1600 usually adds some processing for noise reduction, but the file size is also lower (likely due to lower noise and maybe also being very underexposed) so that could be a factor too.
I agree test 4 is a bit slower than the equivalent L shots. This matches previous experience that resizing has some performance penalty. In test 5 (M2) the smaller amount of data might compensate for the overhead of resizing. But in any case, the differences aren't large. IIRC, the impact of resizing was much larger on Digic II cameras.
So overall conclusion is none of this make much difference
on this camera. I wouldn't assume that A2500 behaves exactly the same.
