I start with two premises: Premise 1: All human beings are bad at philosophy. Premise 2: Kant was a human being. Therefore, um, uh, let's see....
It is sufficient a person's for being "bad at philosophy" in the relevant sense that when that person tries to build an ambitious, elaborate philosophical system that addresses the great, enduring questions of metaphysics and epistemology, there will be some serious errors in the system, as a result of the person's cognitive shortcomings (e.g., invisible presuppositions, equivocal arguments). It is very easy to be bad at philosophy in this sense, and we have excellent empirical evidence for Premise 1. Premise 2 also seems well attested. Further supporting evidence for the conclusion comes from the boneheaded things Kant sometimes says when he is speaking clearly and concretely rather than in a difficult-to-evaluate haze of abstracta.
Here's a vision, then, of Kant:
Kant has a brilliant sense of what it would be very cool to be able to prove -- or at least a brilliant sense of what lots of philosophers think it would be very cool to be able to prove. For example, it would be very cool to (non-question-beggingly) prove that the external world exists. It would be very cool to prove that immorality is irrational. Kant also has some clever and creative pieces of argumentation that seem like promising elements in potential proofs of this sort. And finally, Kant has an intimidating aura of authority. He creates a fog of jargon through which the pieces of argument appealingly glint, in their coolness and cleverosity. And, voila, he asserts success. If you fail to understand, the fault seems to be yours.
Maybe this sounds bad. But the thing is: There really are interesting pieces of argument in there! It's just that they don't all fit together. There are gaps in the arguments, and seeming inconsistencies, and different possibilities for the meaning of the jargon. Because these gaps, seeming inconsistencies, and possibilities might be variously resolved, there need be no one right interpretation of Kant. We can be Kant interpretation pluralists. Although there are clearly bad ways of reading Kant (e.g., as an unreconstructed Lockean), there might be no determinately best way, but rather a variety of attractive ways with competing costs and benefits.
Interpret the terms this way and fill in the gaps that way, and find a Kant who thinks that there's stuff out there independent of our minds that causes our sensory experiences. Interpret the terms this other way and fill in the gaps that other way, and find a Kant who regards such stuff as merely an invention of our minds. Yet another Kant holds that there might be such stuff, but we can't prove that there is. Call these Kant Model 1, Kant Model 2, and Kant Model 3. There will also be Kant Model 4, Kant Model 1a, Kant Model 5f, etc. Similarly across the range of Kantian issues.
But surely only one of these things is what Kant really thought? No, I wouldn't be sure of that at all! When our terms admit multiple interpretations, when our arguments are gappy and dispositions unstable, the contents of both our occurrent thoughts and our dispositional opinions can be muddy. When I say, "the only really important thing is to be happy" or "all men are created equal", what exactly to do I mean? There might be no exactness about it! (See my dispositional approach to attitudes.) This is as true of philosophers as of anyone else -- and, I would argue, as true of the mortal Kant as of any other philosopher.
But even if Kant did have absolutely specific private opinions on all the topics of his writings, it doesn't matter. The philosophy of Kant is not that. Maybe in the secret grotto of his soul he was an orthodox Thomist and he invented the critical philosophy only as a joke to amuse his manservant Martin Lampe. This would not render the Critique of Pure Reason a defense of Thomism. Kant's philosophy is embodied in the words he left behind, not in his private opinions about those words. And those words might not, very likely do not, determinately resolve into one single self-consistent philosophical system.
Historians of philosophy can and should fight about whether to treat Kant Model 2b, Kant Model 5f, or instead some other Kant, as the canonical Kant. But those of us who don't make Kant interpretation our profession should have some liberty to choose among the Kants, as best suits our philosophical purposes -- as long as we bear in mind that Kant Model 2b is no more the One Kant than Hamlet Interpretation 2b is the One Hamlet.