I know that for earth threatening objects, the probability of impact is calculated taking a draw from a population representing the supposed distribution of the orbital elements, propagating it and seeing whether or not it impacts the earth. This process is repeated literally millions and millions of times and the probability of impact is the relative frequency of hits.
If they applied similar techniques to this asteroid, they may have seen that the draws that result in an impact (and there would be a lot of them) are clustered around a certain area of Mars, let’s call it “the impact zone”. If none of them were anywhere near Rover, then they might, not altogether unreasonably, conclude that “Rover is outside the impact zone.”
They know very closely when the asteroid will pass through the target ellipse. The Rover would be on the side away from the impact if it happens. The unknown is how far debris might scatter, if it might somehow go around the planet and hit the Rover anyway.