No DMont and Gibbs arent "positively correlated" nor are any RBs. Some of the ceiling outcomes include the other guy getting a early injury giving the other back all the work. If one of the RBs is mashing down impressive runs the OC is more prone to keep giving him the ball. Just like if one back fumbles once or twice it will provide the other guy a easier route to a gpp winning outcome.
The better the offense performs, the more they score, the more plays they run, the pie they share gets bigger, and they are going to have around the same percentage of carries regardless, so yes, they are, in fact if correlation data you're using doesn't have that as the case, its clearly incorrect. The upside case, you're kind of right, but only because of price.