It seems x, the inclination from the equator in the AAK model, is well defined within ]-pi/2,pi/2[. Should probably throw an error outside of this range. Current sanity check in the base class doesn't account for retrograde orbits not being well defined.