A point source of light isn't likely to straddle more than one of the cell induced distortions on the mirror's surface. However, the distortions can change the size and apparent displacement of the point source of light relative to the rest of the mirror.  In making an orbital determination at least 3 time separated exposures are required. The length of the chord separating the first two positions is comparred to the chord length of the last two positions where the middle position is common to both chords. When the two chords are equall in length the orbit is circular, if one chord is longer than the other, the orbit is elliptical. Here the difference in chord lengths serves as a clue to the orbit's eccentricity. If one or all three of the exposures places the point source of light over differently distorted regions of the mirror, the orbital calculations will be skewed.
At these distances, tiny changes represent very large changes in what is being observed, so what may be an elliptical orbit that will cross Earth's path could appear to be a circular orbit that will never cross Earth's path.
I began this paper with the question: Can our modern engineering practices of only making things good enough to do the job, interfere with our detection of dangerous cosmic missiles? The answere is a subtle yes.  Subtle because we have no problem detecting the missiles, we just don't get accurate enough measurements to recognize all the potential colliders among them.
Just how threatening can an 20 km asteroid be?  No threat at all as long as its orbit keeps it out of our atmoshpere.  Travelling through our atmosphere at a velocity of 20 km/sec it would heat up to the point where it explodes with with a force of 7.675 Million Megatons of TNT. Since 20 km/sec is well below the Earth's escape velocity, if the asteroid didn't explode on the first pass through our atmosphere, it would likely do so hours later on its second pass through our atmosphere, as it would then have a small highly elliptical orbit around the Earth.