Upload & Sell: Off
How about the lowest ISO you can get away with in whatever lighting condition you're shooting.
It's funny, I just spent the last hour looking at different iso stops in pictures I've taken with my 5D3 (now sold) and now I find this thread. Its quite evident to me that 160, 320, and 640 are the best iso's through 800. Beyond that it seems to make no difference. ISO 100 has more evident banding than and blotchiness in the shadows than 160, while 125 is the biggest offender. The same holds true 1 stop up from each of these 1/3 stops. Is it a huge difference? No. But if you feel you may be manipulating the shadows at all, stick to 160, 320, 640 and definitely stay away from 125, 250, and 500.
These are my findings after looking at my pics. I believe the same holds true for the 5D2
I've shot a 5D2 for a bit more than four years, making tens of thousands of exposures, and producing many prints in sizes up to 24" x 36". This whole "fractional ISO" business impresses me as measurebation of the most bizarre type. Here's the deal. I could show you two 24" x 36" prints of extremely high quality produced from images shot on the 5D2, with one exposed at ISO 100 and one at ISO 200, and you would be completely unable to tell which is which. I could likely include one shot at ISO 400 and absolutely no one could tell which it was unless they were challenged to figure it out, looked extremely closely at the prints side by side... and then they might have a sort of hunch ("is it this one maybe?") as opposed to seeing a clear and definitive difference. Heck, I could even toss in a very fine ISO 800 shot that had been carefully and appropriately post-processed and you would be very impressed with the quality... and it wouldn't occur to you to wonder about the ISO.
Of all that things that might actually make some difference in one's photographs, aesthetically or technically, worrying about which partial ISO value to select for optimal image quality is... not even on the list.
Would really like to know how much difference these measured artifacts make in real world photography. Also Dan Mitchel's observation that part ISO stops aren't really important is borne out by these charts. Yes, there are inflections between whole numbers but they are unlikely to be statistically different.
Or, more importantly, unlikely to be visibly different. ;-)