If you're all the way zoomed to the tele end, it doesn't matter how the W setting for AFMA is set, the camera uses only the T setting. Same when zoomed to the wide end, the T setting is ignored. So, if you're just testing at the ends, order doesn't matter. For any zoom setting that's not at the ends, the camera does a simple linear regression between the W and T values (even if one is set to zero). If you want to test an in-between focal length (on the 24-105, I test at 24, 50, 85, and 105mm, to get an idea of performance across the range, and I test each FL at 25x and 50x the FL from the target), you must set both W and T to the same AFMA value.