There is a tension in the Critique of Pure Reason between the novelty of transcendental idealism, on the one hand, and the terms and ideas of 18th century philosophy, on the other. Kant was, for all his importance in the subsequent development of philosophy, and for all his remarkable originality, still bound by his place and time. He could not shake off the remnants of the old rationalism, though the First Critique certainly laid the groundwork for its complete destruction by the post-Kantians. In a sense, of course, the First Critique was intended not as a radically new vision of philosophy but as an answer to a simple problem - how can knowledge be justified in light of Hume? A patching-up of epistemology, as it turned out, would not do, but it is useful to keep in mind that Kant himself had at least that modest goal in 1781.
There is, as I said, a tension between the new and the old, between a radically different conception of reality and between a modified rationalism/empiricism hybrid as the end result of the critical project. Kant seems as times not quite willing to make the idealist leap and to shut off the noumenal utterly from human knowledge. Thus, in the Refutation of Idealism, real external objects are actually known to exist, not just assumed as placeholders in a vague metaphysics. But if transcendental idealism is to serve fully in purging thought of unnecessary assumptions, it seems positively beneficial to deny real knowledge of independent existences. We can otherwise never be sure that our experience is of properties of the thing itself or properties imputed to things through the mind's modification of objects for experience. By acknowledging that all experience is mind-dependent, transcendental idealism gives up the impossible goal of real knowledge and settles for knowing the mind's forms of thought themselves, confident that an inventory of concepts will serve as the only possible substitute for a mystical metaphysics of noumena. This assumption of a complete break between reality and experience is disquieting to some because it seems to imply throwing out the touchstone of truth (objectivity) for radical subjectivism. In fact, the whole purpose of the First Critique was to show how this so-called "subjectivism" is in the fact the only way of vindicating objectivity - we can be certain of those aspects of experience that we ourselves contribute, because those forms of thought are universal and necessary, and what is universal and necessary cannot be false. Against Hume's objections, nothing better seems to have been raised.
Transcendental idealism could serve two purposes. It could provide a critique of the grounds of knowledge, allowing us to see what we contribute to experience and what an independent reality contributes, but allowing room for direct knowledge of things through some faculty other than cognition. On the other hand (and this path seems to have been taken by Hegel, among others), transcendental idealism can serve as executioner of the rationalism/empiricism dichotomy and as propaedeutic to a new way of thinking that locates the reality of experience in thought itself. Even this radical idealism can leave room for the existence of mind-independent things, but it utterly rejects possible knowledge of them. In other words, mind-independent things could exist, although we could never know how or if they do, and, in any case, they are completely irrelevant to theory and practice in all fields.
The radical second path is more consistent with the critical project. Once Kant has made the ideality of space and time (and thus of all reality that can be presented to the senses, or that can be considered as a possible object of the senses [i.e., geometry and arithmetic]) seem self-evident, and once he has gone on to locate the basic elements of physics in the mind, not in the (independent) world, the question "How can knowledge be objective if we contribute to its creation?" is turned on its head; instead, it becomes difficult to fathom how knowledge of independent things is possible or could be known with certainty. If the mind must change things in order to understand them, mind-independent knowledge is incoherent. Coming at transcendental idealism in the 21st century, we can criticize Kant more freely for not making the leap. Once Kant himself introduced mind-dependence in such stark fashion to the intellectual world, it quickly (although not immediately) became impossible to discuss knowledge without assuming something like transcendental idealism going on in every branch of it (though such a revision of metalogic was rather late in coming; still, it appears to be here now, at least). Thus the nonsense view now is that real knowledge is possible; before Kant, the nonsense view was that knowledge was still knowledge after being filtered by subjective demands.
Of course, the discussion above hardly gets into what Kant seems not to have considered - whether mind and experience exist in a reciprocal relation of modification (Hegel would dare to say "clarification"). It took a further development in idealism to propose that the instrument of knowledge could change itself, or be changed by reality. Although experience, for Kant, may be a constant product of cognitive organization, the forms of thought were fixed - thus objectivity. Saving objectivity in light of Hegel is, perhaps, an impossible task; it is left as an exercise to the reader (!).