There seem to be a lot of questions that pop up here about failed applications of the Categorial Imperative, somehow "disproving" its validity. This seems misguided. Since I am not a Kant expert, I hope others will point out any misunderstandings on my part.
This attention to applications is, I believe, somewhat misplaced and stems from the difficulty we have today in grasping philosophical idealism in general and the modern transcendental variants. We are simply too steeped in physics, utilitarianism, and psychology.
Kant's CI might best be understood as something like Newton's laws of motion, though of course as the "laws" of reflective, mental motion they can hardly be applied with mathematical precision. The CI is an a priori principle that is necessary to beings endowed with reason, freedom, and moral sensibilities. One can reject that description entirely in favor of selfish genes, or whatever, and in that sense not only the CI but all of Kant's project can be deemed a "failure" of precise physical description.
But if we tend to accept that, yes, we have a capacity for reason, autonomy, and moral duty--as well as a capacity for scientific knowledge of the world--then Kant is demonstrating the a priori structure necessary for the coherent integration of such capacities. One essential part of this structure is the highly general principle of the CI. Just as Newton observed "motions" and derived a universal principle within a coordinate system, so Kant observed "emotions" and moral behaviors in the phenomenal world and derived the abstract, noumenal principle to which they all refer and relative to which we may understand them.
Even Newton's laws of motion, while "universal" and "necessary" in some sense are idealized. They do not apply at certain scales, as we now know...and, crucially, they do not even apply exactly with absolute accuracy in any given, particular instance.They are "regulative ideals" to use one of Kant's terms. They posit an ideal motionless space full of infinite ideal points, relative to which we can plot the motions of macro-objects to a high but technically limited degree of accuracy.
Similarly, the CI is the logical structure derived, we might say, from reflection upon all moral actions in the world, also within the posited space-time limits of "sensibility." Or such is Kant's argument, a compelling one, I believe. We can never know by simple observation whether any particular action is moral by Kant's definition, because we can never observe precisely the intentions motivating it, even in ourselves. Nor can the CI simply be applied in any particular case. If it could be, we might as well replace juries and judges with computers.
We refer to it only as an a priori principle and regulative ideal that helps us explain what we mean by immorality and moral duty as practices in the highly conditional, imperfectly grasped world. It is the logic discernible behind many valid moral and religious practices, and our good "common sense," yet it was important to Kant that it not be based simply on a reference to God or any other authority. In keeping with the spirit of Aufklarung, it is derived critically from the necessary coherent interactions of freedom and reason.
Kant himself states somewhere that he does not expect the CI to apply exactly to actual cases, though I do not recall the reference. In the four examples he gives, he is assuming we accept the ordinary understanding that they are right or wrong, then he asks us to discern what general principle or form they might share in common. How can we express concisely and consistently what makes these various actions appear to us as "moral imperatives"? What logical common denominator describes how the rational being judged some practice to be "right"? That is the CI.
While applications might always be improved, that is an adjustment of judgements within the contingencies of the world, not an "improvement" of the CI itself. And, as with Newton's laws, we may find ourselves adding subsidiary principles as new social situations and even new forms of "common sense" arise, though probably without Kant's blessings. For example, Rawls' "difference principle" seems to me a valuable addition within the contingencies of capitalist democracies, so far from the sleepy civilities of the Prussian order.