For the truth is, that the idea of penal sanction, which is the essence of law, enters not only into the conception of injustice, but into that of any kind of wrong. We do not call anything wrong, unless we mean to imply that a person ought to be punished in some way or other for doing it; if not by law, by the opinion of his fellow-creatures; if not by opinion, by the reproaches of his own conscience. This seems the real turning point of the distinction between morality and simple expediency. It is a part of the notion of Duty in every one of its forms, that a person may rightfully be compelled to fulfil it. Duty is a thing which may be exacted from a person, as one exacts a debt. Unless we think that it may be exacted from him, we do not call it his duty.
So what distinguishes morality from expediency (or, indeed, worthiness, since Mill goes on to add that -- I just noticed that this parallels the System of Logic division) on Mill's view is that morality is doubly supported by utility: there is the support that comes from recognizing that the action or rule is conducive to the greatest happiness for the greatest number, and there is the higher-order support that comes from recognizing that it is conducive to the greatest happiness for the greatest number to attach sanction to it. In a sense we can say that utility governs every sort of 'ought' and 'should'; but there are many kinds of 'ought' and 'should'. I ought not take the 1L bus to get home; this is a practical judgment, and thus a rough judgment of utility. It is even a fairly good judgment of utility, because as a rule I shouldn't take the 1L to get home. But this is not a moral 'should'; and on Mill's view the difference is that it would not be conducive to the greatest happiness for the greatest number for anyone to regard failure to conform to this judgment as punishment-worthy. It is, on the contrary, entirely conducive to the greatest utility for us to consider deviation from the judgment 'I should not murder' as punishment-worthy.
This is related, I think, to two key features of Mill's utilitarianism, features that might at first glance seem to be in tension: (1) the fact that it places such immense emphasis on the importance and (relative) inviolability of good moral rules; and (2) the fact that it has so much tolerance for nonmaximizing, particularly in the forms of practical approximation, rules of thumb, and toleration of bad judgment. On Mill's view, if you have determined by utilitarian analysis that X is the best alternative, it does not immediately follow that it's wrong to do something other than X. To get that judgment we have to engage in a higher-order utilitarian analysis of whether we should regard failure to do X as deserving of punishment. If we don't have that, we've merely determined that it's better for me to do X than not; but simply having determined that doesn't tell us much about right and wrong, any more than I have learned anything about right and wrong from learning that it is better for me to take the 1M than the 1L to get home. If I take the 1L instead, that is not the best way to go, but I can still do something to bring me around to where I need to be; nothing of fundamental importance to society hangs on my being efficient or inefficient, competent or incompetent, on a matter like using the bus system to get home, despite the fact that one is better at getting me to my destination than the other. Similarly, if A is more conducive to the greatest happiness for the greatest number than B, it doesn't follow that I must do A; it could well be that B, although not as good as A, is good enough, and that in both ways I can work toward the greatest good for the greatest number, although in one way more efficiently than in the other, and perhaps in one way get a better result than in the other. For it to be the case that I must do A, I must show that the pursuit of the greatest happiness for the greatest number cannot tolerate my doing anything other than A. Not everything conducive to the greatest good for the greatest number is equally conducive to it; some are slightly conducive to it, and these are expedient, but some are such that without them the effectiveness of my pursuit of utility goes almost to nil, or even worse, my pursuit becomes actively destructive of our ability to pursue utility, and these are obligatory.
Thus it is that on Mill's system moral rules, to the extent that they are more or less good judgments of utility, override almost every other utilitarian consideration; they are multiply protected. In order even to show that it would be better not to follow a moral rule in a given case, I would have to show both that in that case it would be contrary to utility to consider violation of the rule as deserving of punishment and that following the rule is more conducive to utility than the opposite in that particular case. If I only do the latter, I haven't undone the moral rule: the moral rule is still supported by its utility-derived sanction. If I only do the former, the moral rule is no longer obligatory, but it is still the better thing to do. And this, mind you, is just in showing that it would be better not to follow the rule in that particular case; even if I did both of these, it would still be permissible to follow the moral rule. To make it so that it is not permissible to follow a moral rule in a given case, I would have to show that utilitarian considerations require that following the moral rule in that particularly case be deserving of punishment. The circumstances under which well-established moral rules would not swamp all rival considerations would have to be very peculiar. It seems clear that such cases can indeed arise on Mill's view; moral principles presumably get refined by our discovery of new circumstances of this kind, which then leads us to modify the formulation of the principle so that it properly covers even these circumstances. But for the most refined versions of the most well-founded principles, one can in any ordinary circumstance regard them as virtually iron-clad.
Thus we find that by giving an account of how both morality and expediency can both follow as distinct departments from utility that Mill can both insist upon the immense and (for most practical purposes) total superiority of morality over considerations of policy and tolerate a great deal of deviation from the ideal pursuit of the greatest happiness for the greatest number. And both of these, I would suggest, are actually very important to Mill's project of building a liberal society, since such a society requires recognizing the supremacy of considerations of justice over considerations of mere policy but also requires allowing a great deal of deviation from perfection. That makes it an interesting account, since one might have thought (and people did think, since Mill has to deal with arguments that are based on each) that both of these would give the Millian utilitarian some trouble. The moral positivism of making morality to be based on sanction, resulting in what is basically a secularized divine command theory, is not something appeals to me; but I find it interesting how Mill manages to handle so much with such a simple account of such a simple distinction.