The point that the author is trying to make is that there exist some circumstances under which non-moral actions might be viewed as admirable. Obviously, one cannot pursue this further without stating one's ethical beliefs or atleast elucidating what he/she means by "moral". The author says that this proposition is valid even in consequentialist theories. But, this can only hold for objective consequentialist theories.
This does not hold in subjective consequentialist theories, as can be easily seen because, by definition, what is called "moral" in such theories is what brings about the best consequences. A non-moral act under this framework is one that does not bring about the best possible consequence.
If one is to make the reasonable assumption that what is "admirable" (atleast in the sense of "admirable immorality") is what is "good", it is easy enough to see that admirable immorality is not a possiblity.
The author of the article says, in conclusion, that admirable immorality is possible when:
(1) There is a moral code: that is, a socially accepted set of instructions for bringing about outcomes that are justifiable in terms of considerations that have moral weight.
(2) Actions that go against the dictates of the moral code are considered immoral.
(3) Due to the complexity of the moral universe, and to pragmatic limitations on the nature and complexity of moral codes, the moral code occasionally dictates action that fails to bring about the best possible outcome, as judged from the perspective of those considerations that have moral weight.
In this, the author is justified. But, it is also worth mentioning that AI is not possible when what is "moral" is deemed to be what brings about the best consequence.