You run a 65 bit range searching for pubkey Z, solve in expected time, and accumulate x amount of DPs to solve (whatever number you want to use, it is irrelevant)
Now, you tamed the wilds from the previous 65 bit run.
Lastly, you run a 65 bit range searching for pubkey W, on average, how much quicker will you solve pubkey W versus pubkey Z?
A LOT BETTER, if this was your question. And every time I benchmark to test for a new solution, I tame the found wild DPs, so results improve themselves. I also trim periodically DPs that are outside of some range, since almost 99% of the solutions are found via a DP which is an specific median range. I almost never find a key by hitting a DP which is below 50% of the interval, for example, even though there were a lot of them in that region. Don't ask for why that happens, I can only speculate it has to do with average travel size, who knows.
Currently I have a 80-bits solver DP database that can solve any key in under 6 seconds on a GPU, and under a minute if using the CPU.