Remember assumption 3: tanq @ q, so the paraxial ray angle is replaced by the ray slope. Now consider this macro: 

This gives the result:

As the marginal ray angle set by the solve goes to zero, the paraxial marginal ray angle and tangent tend to each other. At this large marginal ray angle however, they are quite different.

To trace parabasal rays, trace a real ray very close to the chief ray and scale it to the desired pupil coordinate:

This gives

It can be seen that it is the parabasal marginal ray angle, or tangent of the paraxial marginal ray angle, that is set by the solve.