The main reason is that the other two major factors in performance, optical design and optical fabrication, have already been improved to the very limits of what can be accomplished. After spending good money on a near perfect design and fabrication of optical components does it make any sense not to assemble them in perfect alignment?
By perfect alignment we mean placing the elements where the design says they should in the theoretical optical and mechanical design, not just within some tolerance band, but precisely where the design says they should be. Only this way can you get performance at the same level as the perfection of the design and fabrication of the components.
In the last several decades with the help of incredible computing power optical design has reached the limits of perfection. Optical systems are being designed with vastly better performance than ever before by making use of aspheres and coatings that could only be used because of the computing power available these days. Unfortunately these advanced designs could not be made because optical fabrication methods had not kept up with the power of optical design.
Now with computer controlled polishing techniques like MRF, diamond turning and molding of both glass and plastic elements the aspheres that help with performance because of their design are now practical to make to high quality. Similarly coatings to enhance performance have been improved to the point that making them any better is no longer cost effective.
Now comes the hard part; how do you position these near perfect components, mechanical and optical, relative to each other so they perform as well as the theoretically perfect design? One way, the way usually followed is to tighten the tolerances on both the optical and mechanical parts until they have to go together perfectly but this is impossible to do because some slop has to be left or it is impossible to get the glass into the metal mount.
Another way is to leave loose tolerances for edging and bores of lens barrels so the element can be positioned where the design specifies it should be. This is where a tool like the PSM comes in. The PSM can be positioned so its focus is where the center of curvature of a surface should be according to the design and then the surface adjusted until it is centered to the PSM reference cross hairs.
Clearly this is not the way to assemble optics on a mass production basis. But when performance is at a premium such as in lenses for reconnaissance and cinematography, for example, it is far more efficient and cost effective to allow looser tolerances on glass and metal, and then individually align each element as it should be.
I remember many years ago seeing sophisticated lens systems assembled from well-made glass and metal components which were then tested for optical performance. They often failed the optical test and were sent back to the assembly department where they were taken completely apart and re-assembled with the hope that this time they might pass the performance test. In hindsight this was a ridiculous procedure. With the aid a a PSM and a little Fixturing the lens systems could have been assembled so they passed the optical test the first time, every time. This is the smart way to do alignment.