So I did some testing regarding attack buffs in quest 425. I used illustrated Yuffie (speed) on Ava (magic) on Starlight +21 in the 3rd and 4th slots.
3rd slot (no keyblade boost), each values is an averages of 11 measurements
Speed:
lv.1 -1.17
lv.2 -1.38
lv.3 -1.64
Neutral:
lv.1 -1.39
lv.2 -1.70
lv.3 -1.98
That's right, color specific boosts suck for medals that receive no keyblade boost.
4th slot (keyblade boost), each values is an averages of 2 measurements ( I was getting bored)
Speed:
lv.1 -1.15
lv.2 -1.39
lv.3 -1.64
Neutral
lv.1 -1.31
lv.2 -1.55
lv.3 -1.78
Color specific values remain pretty much the same but the neutral boost now provides a lower but still much better multiplier. I then checked different combination of neutral and color specific boosts and the values do indeed multiply when calculating damage. Some tips for combining neutral and color specific boosts for medals that receive keyblade boosts:
lv. 2 neutral > lv. 1 neutral + lv. 1 color
lv. 1 neutral + lv. 2 color > lv. 3 neutral ~ lv.2 neutral + lv. 1 color
lv.2 neutral + lv.2 color > lv.1 neutral + lv.3 color > lv.3 neutral + lv. 1 color
But the fun doesn't end here. I used illustrated Cid (power) to test how a not-effective medal would fare. Averages of 2 values.
3rd slot (keyblade boost)
Power
lv.1 -1.35
Neutral
lv.1 -1.28
The neutral boost multiplier stays roughly the same as Yuffie's but the color specific one gets a huge increase. A neutral effective medal might have yet another set of multipliers but I'll leave that to someone else. I have had enough number crunching for now. Oh, and the RNG can have a pretty big effect on damage output: I have observed +/- 0.04 on the multiplier, which is ~2-3%. Think about that the next time the enemy has a sliver of health left and you fail to one turn a quest