It's conceivable that in the distant future we will have simulations indistinguishable from reality, but there is no way to measure the probability that our world is a simulation. It's not that there exists a way to measure and we are just uncertain about the constants that go into the calculations; rather, given the assumptions of Bostrom, and assuming infinitely many worlds (potential infinities suffice), you reach a contradiction.
It comes down to the same problem inherent in saying "pick a random number" while assuming all numbers are equiprobable.
You must weight the numbers differently in order to have a probability measure.
Part of the problem here is trying to put this into a probabilistic framework. You can probably still say something useful without trying to invoke probabilities. You can probably define some weaker notion of "occurrability".