The errors should be a function of time until impact. I'd like to know what that error variance vs time plot looks like.