I'm not a GD&T expert but it would appear to me that if you have already toleranced the position of each hole, then variation in the distance between them is already set by that. Each hole can be off a maximum of a certain amount in a particular direction, and you can total that variation to come up with the overall distance tolerance.
However, you might also want to think about it from a statistically based tolerance stackup perspective. What is the probability of being off by a certain amount for each hole, and how does that stack up. To me it would not simply be the mathematical sum of the two tolerances, but some amount less.