The test that I read was very through. Weighed, sized and primed cases were charged with powder that was from a freshly opened cannister. The charge weights were recorded (total weight minus weight of each primed case), and some were chronographed. The open cases of powder were racked in a closed container with a significant amount of desiccant below and left that way for some time, until their weights were as low as thy would go, then the cases of powder were weighed, and some were chronographed. Then the remaining cases were racked in a closed container with water below them and left till their weight stabilized, and some were chronographed. The difference in charge weight from dry to wet was 2%, and the difference in velocity from dry (highest) to fully humidified was I believe 320 fps. I believe that this sort of test trumps theory and supposition every time.
Another example comes from Jim Borden, who loaded a number of batches of ammunition for a large match, over a period of a couple of weeks (or more, I forget) in which the weather varied considerably. There were wide variations from session to session, in the heat and humidity. At the start of each session he adjusted his load so that it produced the desired velocity, because he views tune as velocity specific. By the time he had, I think 300 rounds loaded, the extreme spread, batch to batch of charge weights required to produce the same velocity was 1 1/2 grains. His shooting buddy, told him that he had made a mistake in his approach that would surely lead to accuracy problems at the match, so some rounds were pulled from each batch, and groups shot with mixed ammunition. The groups were pleasingly small. It has been a while since I was told the story, but I am pretty sure about all of the important details, especially the grain and a half, and the accuracy obtained when testing.