A circular orbit requires very precise initial conditions; an elliptical orbit does not.
To see what I mean: Imagine a rock which is a certain distance "R" away from the sun. It is your job to "push" the rock with a speed and direction of your choice, and see what happens.
It turns out that, in order to set the rock into a circular orbit, you must (a) push it with a very particular speed (which is a function of "R" and the mass of the sun); and (b) push it in a direction which is exactly perpendicular to the line joining the rock and the sun.
On the other hand, say you give the rock an initial push in some RANDOM speed and direction (assumed different from the "perfect" speed and direction required for a circular orbit). In this case, the outcome is determined only by the speed of the push; and in particular:
* If the speed is less than some critical speed "Ve" (which is a function of "R" and the sun's mass), then the rock will go into an elliptical orbit;
* If the speed is greater than "Ve", then the rock will go into a "hyperbolic" trajectory (it will escape from the sun's influence rather than orbiting it).
So this means that, when a random collision alters the path of an orbiting object, it is highly improbable that the result will set the object in the precise speed and direction required for a circular orbit. It is much more probable that it will set it into one of the (infinite range of) speeds and directions that will result in an elliptical orbit.