My thoughts were not to use a #2 filter 'specifically' or to rely on the grades being ANSI.

I just think it would be a good idea to use a filter (or filtration) instead of unfiltered so that you don't have to figure out the next step.

Another thought. It might be good to make first round tests on whole filter number jumps instead of half filters.

It goes back to the f/stop time idea that you work up tests in dramatic increments so you can see the potential and know your direction. Then refine it once you know it has to be in-between.