They are very different. In what follows I shall describe three salient discrepencies. The differences are so drastic that even people who are very familiar his works feel surprising.
Let me start with disjunctivism. McDowell makes clear that his disjunctive conception is epistemological, but he seldom says explicitly about his attitudes towards other versions of disjunctivism. He grants that the good case and the bad case can be phenomenally indistinguishable, so presumably he rejects phenomenal disjunctivism. How about the version concerning state? He is not clear about this, but some passages in "Sungular Though and the Extent of Inner Space" can be read as proposing state disjunctivism. Now the most controversial one is the content version. In defending "object-dependent" thought and de re sense, he seems to commit that the good case and the bad case cannot share the same kind of content. However, in discussions he said that they can. I am sure that he said this in Taipei, London, and Canberra. Many people are surprised by this point, including his former students and experts of disjunctivism. I think he is drawing the distinction between constitution and entertain: object-dependent thoughts constitutively depend on acqaintance of relevant objects, but after the constitution, those thoughts can appear in bad cases. Consider the Twin Earth Scenario: the idea of "water" constitutively depends on water substances, but this does not imply that whenever we think about water, there must be water around.
The second is about the idea of the conceptual. McDowell holds that the rational and the conceptual are co-extensive, and this co-extensiveness is entirely stipulative. He is explicit about "conceptual -> rational," but not clear about the other direction. I think the other direction should be a substantial claim (as opposed to stipulative), or the co-extensiveness will be empty, and he does not really need to take issue with non-conceptualists. However, in discussions he insisted that the entire co-extensiveness is stipulative, and this is not a change of mind. I find this hard to be reconciled with many passages in print.
Finally, the relations between the following three distinctions: the space of reasons/the realm of law, second nature/first nature, and world/environment. McDowell writes that mature human beings are in the space of reasons (by being initiated into second nature), so they can have the world in view. This seems to imply that the other side of those distinctions also go hand in hand, i.e. the realm of law, first nature, and environment are the so-called "physical world." But in discussions McDowell refused to identify environment with physical world. I think this move is odd. Do we want our picture to have physical world/environment/ world in McDowell's sense? This three-layer structure seems to be too excessive. I think we can still identify environment with physical world, if we understand the latter as containing affordances, solicitations, and even reasons (though lower animals cannot be responsive to reasons as such).
The general difficulty is this. When we write about McDowell, we need to refer to specific passages. But McDowell in person and in print are so different. Besides, to cite conversations with philosophers is a bad practice. It is not clear to me how not to misinterpret McDowell if we take his works in print at face value. But anyway, it is still fruitful to engage McDowell. I think I learned very much from the discussions in Australia in the past two weeks. Special thanks to Huw Price and David Chalmers for organizing the two conferences.