The variation in results by color is not surprising. If you recall, each wavelength of light has a slightly different focus. If you were to only use one wavelength - IR - there is a significant adjustment. Many of the older lenses have a mark on the lens to indicate the adjustment. IR conversions provide the ability to adjust focus so that it is corrected for IR spectrum instead of the visible spectrum. Red spectrum adjoins IR spectrum, and is the most likely to show a focus shift.
Likewise I would not expect the same AF fine tuning value for different sensors. The purpose of AF fine tuning is to adjust the focus of the lens camera combination. Normally the center is the most accurate sensor and the most accurate portion of the lens. Expecting the outer sensors to have the same AF adjustment as the center adds a number of variables beyond just the camera body combination.
Variation in the outer sensors can be a function of lens design as well as lens issues or defects. Most lenses have less accurate focus at the outer edges compared to the center, so even if the AF sensor is accurate, the lens is not as sharp.
Add to that a difference in the AF performance of sensors. The center sensor is different than the cross sensors, and they are different still from the non-cross sensors. The center sensor will AF at f/8, the cross sensors at f/5.6, and the other sensors at f/4. Tracking and AF speed are considerably slower and less reliable with outer sensors. Focus errors are greater with outer sensors. And frequency of accurate focus is better with the center sensor than outer sensors.
AF systems are getting better and better each year, but they are not perfect. A recent test comparing the D4, D800, and D600 AF tracking showed 80% successful captures with the D4 and only 45% with the D600 - and the D800 in between. But 45% was still far better than manual focus for action shots by anyone but the very top pros.
Zoom lenses perform differently at different focal lengths. I saw test results recently where the greatest error on a 24-70 lens was around 50mm - and it was in the opposite direction of the errors at 24mm and 70mm.
Take a look at some of the articles of focus and testing at LensRentals.com . Roger Cicala has done a nice job of testing, but he does not even bother to test the extreme sensors for AF fine tuning or AF consistency and accuracy. He also has invested lots of money in test setups to accurately fine tune and test for specific issues.
As Linwood suggests, this is all about compromises. If you really want the best result, you can send your lenses with the camera body for service. If you truly want perfect results, you are probably joining the camp of Lloyd Chambers and only shooting with Zeiss or Coastal Optics prime lenses that are manually focused using Live View.