The purpose of testing is to eliminate as many variables as possible so that is repeatable from one test to the next. Are you going to make a dyno pull, make a tuning adjustment, then drive to a different dyno for the next pull/adjustment? No, that's stupid. But in a way that's what you're doing with a vdyno.
The nature of a virtual dyno is going to introduce a bunch of variables that you have no control over. You could absolutely use the same stretch of road every time, but a simple change in wind direction, temperature, or surface adhesion (caused by road grime, moisture changes, oils, etc), among other factors could effect your results.
I'm not saying that it isn't an effective tool, it can be. I was simply saying that it is introducing a lot of variables or items that are left up to the user which is best avoided whenever possible.