He is referring to a Pass through mode and even non CA corrected prisms do not show CA.
No he's not actually. The ideal beam for any anamorphic system has parallel rays. Projection systems have, by their nature, converging rays, and hence are not ideal for anamorphic systems.
When you put two prisms in an anamorphic configuration they work best when the beam entering the system has rays which are parallel to the optical axis. For example, a diode laser system with inherently parallel rays, but which (due to the nature of laser diodes) natively produces an oval-shaped or elliptical spot, can be made to produce a circular spot with a set of anamorphic prisms used to widen the narrow waist of the spot. This is a common solution to oval-shaped laser beams.
However, when the rays are converging
, that is coming from ray pencils of a definite diameter at the projector end of the beam to a theoretically single point on the screen, the theory of anamorphic prisms is tested, and begins to exhibit aberrations (particularly astigmatism and color aberration). Because the rays entering the anamorphic lens are not parallel, the exit angles are not quite maintained enough to ensure a perfect spot.
In the end this means the larger the throw ratio, the closer to parallel the projector beam rays are, the closer to ideal is the performance of the prism system. This is why prism projection systems worked so well in large cinemas, where the throw ratio would typically be greater than 3.5. In HT applications, with some trying to use TRs of as little as 1.5 and 1.6, prisms perform poorly, as this is a long way from their ideal working environment.
Incidentally, cylindrical lenses designed for theoretically parallel rays perform in a similar manner to prisms in the same parallel-ray environment. However, prisms are fixed in form: four plane surfaces (the angles may vary, but the topology of all prism systems is basically the same). Cylindricals, while they can and often are designed with a "parallel-ray" model, can also be specifically designed for a sharply converging rays typical of HT applications.
In short, cylindrical systems have a better chance of delivering better performance in HT applications than prism systems. The caveat is that this potential
performance advantage is always dependent upon proper design. A poor design of any optical system will not take advantage of any inherent strengths that system may offer on the drawing board.