I use the three point method. I don't see how you have to survey both sides with an indicator when finding out how flat a single side is. If they call flatness on both sides, then I'd agree.
Like when I make a part from 7075 that is 26x18" that starts off at 1.08" thick and ends up at .984", and require .008" flatness on one side, I just set it up on jo blocks and sweep it all over with an indicator. We don't care how flat the other side is (really; it could be off .02", wavy, tapered, etc, doesn't really matter). I do understand that if I do not inspect to find three "high/low" spots on this plate, and instead set it up on three random places, the measure of flatness may say that the part has more deviation from a plane than it really has, but never less.
Imagine throwing a wide banana-shaped part on jo blocks, and having the part supported on one half of the banana. Sweeping this will give the impression that the surface is WAY OUT on one end. Distributing the blocks evenly (both ends of the part) will result in your banana having much less deviation from the same plane.
But anyway, enough of that rambling. I believe that the indicator method would work well. And given the tolerance, I don't see why the feeler method on a surface plate wouldn't work well too, assuming the parts don't come close to being off .02"