I think it's more than coincidence that adopting the gloves coincides with his % dropping back down under 2% once he was healthy.
Yeah, I see the change in the numbers. I just don't agree about what it means.
The thing is, most QB fumbles take place on sacks. Small number of plays overall, but a big impact. When you use attempts in your divisor, it blunts the impact of those sacks so much that you actually lose meaning. If I look at a QB, and I see that he had 35 TDs one year, 10 the next, and 31 the next, do I think he had a terrible season in the middle year? No, I figure he probably got hurt. In the same way, if I see a QB go from 6 fumbles to 12 to 5, rather than figuring he forgot how to hang onto the football in the middle season, I figure he probably was getting sacked more. (Though with numbers that small, small sample size is maybe an even more likely culprit.)
Warner goes from 3.6 to 2.2 to 5.1 to 2.4 to 1.7 to 1.9 in your calculations. You say it's the gloves. I say it's that his sack rate went from 12.3 to 5.8 to 7.7 to 4.2 to 4.2 to 4.5. If most QB fumbles occur on sacks, it just makes sense that any flux in how often a guy is sacked is going to have a big impact in how much he puts the ball on the ground.
When you include attempts in your equation, you won't see that impact.