As I noted earlier, just war theory distinguishes between the principles of jus ad bellum (resort to war) and those of jus in bello (conduct in war). According to the Theory, the latter are independent of the former, in the sense that what it is permissible for a combatant to do in war is unaffected by whether his war is just or unjust. Whatever acts are permissible for those who fight in a just war (“just combatants”) are also permissible for those (“unjust combatants”) who fight for aims that are unjust. Combatants on both sides have the same rights, permissions and liabilities — a view commonly known as the “moral equality of combatants.” According to this view, if we accept that it is permissible for just combatants to participate in warfare, we must also accept that the same is true of unjust combatants. Both just combatants and unjust combatants act impermissibly only if they violate the rules of jus in bello — that is, only if they fight in an impermissible manner.
This has one immediately paradoxical implication: namely, that if unjust combatants fight without violating the rules governing the conduct of war, all their individual acts of war are permissible; yet these individual acts together constitute a war that is unjust and therefore impermissible. But how can a series of individually permissible acts be collectively impermissible?
I think it's actually incorrect that in just war theory jus in bello is independent of jus ad bellum, despite the fact that there are important and morally significant distinctions between the two and despite the fact that the relationship between the two is not always straightforward. The major complications are (1) that combatants, while not usually being the ones who actually decide to initiate belligerent action, have obligations of allegiance, obedience, protection of their fellow citizens, and the like that are not suddenly disrupted and (2) that combatants are not always in a position to fully understand the larger issues in the war -- they can be deceived, they can be misinformed, they can be mistaken, they can be in a position where they are missing crucial information. However, I think there's a more general problem here: permissibility is a very weak modality. Just as something's being possible does not imply that it's actually possible in a given case, so something's being morally permissible does not imply that it's actually morally permissible in the given case -- it just means that there are general considerations that do not of themselves make it impermissible. Take, for instance, McMahan's scenario:
Suppose, for example, that the armies of Aggressia have unjustly invaded and conquered neighboring Benignia. Aggressian soldiers never once violated the principles of jus in bello. But to defeat the Benignian army, it was necessary for them to kill more than a million Benignian soldiers, most of whom were civilians when the invasion began and enlisted in the military only to defend their country from Aggressia. According to the Theory, the only people who have done anything wrong in bringing about this vast slaughter are a handful of Aggressian political leaders who spent the war in their offices and never killed anyone.
In actual fact, there is simply not enough information to go on in this example. What we do know is that the soldiers of Aggressia cannot be accused of acting unjustly simply in terms of the particular format in which individual soldiers carried out the details of the war, since nothing distinguishes their conduct on these particular points from those of Benignia. If there's anything wrong with the actions of Aggressian soldiers, it is not in how they carried out the war, simply considered as such; it is a matter of what they could honestly have known about the war, what options they actually had available, what the situation actually means for their standing obligations as citizens and soldiers, and so forth. These are not usually considered solely in terms of just war theory for the obvious reasons that these are not issues that have specifically to do with warring but with general problems we all face in being members of a society. In fact, many of them will be exactly the same problems faced by non-combatant Aggressians. Just war theory does not rule out that the Aggressian soldiers have done something wrong; there could be any number of ways they are acting wickedly, as human beings, citizens, or even as soldiers generally speaking. To determine this would require much more detail than we are actually given. All that can be ruled out is that any such wrongdoing has to do specifically with the format in which they acted specifically as soldiers in a war situation, precisely considered as such.
A different but related issue arises with McMahan's question above about "But how can a series of individually permissible acts be collectively impermissible?" But there is only a paradox if we are taking permissibility to be a stronger modality than it is. It is entirely possible for something to be permissible in one respect and impermissible in another. For instance, it is entirely permissible for any individual to go without ever having children; if everyone did so, the human race would go extinct; and one could at least argue that it is not permissible for the human race to structure its choices so that the human race goes extinct. But that's not paradoxical; the impermissibility is not any one person's going without children but the whole society's doing so. And that's no more paradoxical than the fact that A, B, C, and D could all each be possible but A & B & C & D could be an impossible combination. Indeed, it is structurally the same; permissibility and possibility are both Lozenge or Diamond modalities, and permissibility is just possibility in some form of deontic modal logic.
This, incidentally, is why I think bioethics often ends up being very screwy; bioethicists tend to argue at the level of the permissible in general. But this is very, very weak; it is entirely possible for something to be permissible in general and yet for it to be actually impermissible in any foreseeable practical circumstances. (Utilitarians are very familiar with this sort of thing; in principle, you can treat anything as morally permissible in utilitarianism, if you only modify the circumstances enough. This is sometimes used to attack utilitarianism. But, while no deontologist would ever accept this as an adequate response, utilitarians will quite reasonably respond that many of these modifications are so extreme that we have no reason to think that they could ever possibly happen in the actual world. If it's a purely abstract and hypothetical question of whether there is some conceivable circumstances under which a grave atrocity would be permissible, that is a different thing from saying there are any reasonably possible circumstances in which there would be nothing at all that makes the atrocity impermissible.) It is generally useless to know that an action is morally permissible at the general level, because this is consistent with its not being permissible in a particular case, or even in all particular cases that anyone will ever actually come across; and permissibility is in practice generally permissibility-in-light-of-the-particular-reasons-that-have-actually-given, and is consistent with the very same action being impermissible for reasons not given. This contrasts with impermissibility, which is almost always useful to know. I often point out to my Intro students that it is only rarely useful to know that an argument is invalid, although it is almost always useful to know that it is valid; there are important differences between this and permissibility (they attach to rather different sorts of things, permissibility is a Diamond modality and invalidity a Diamond-Not modality, etc.), but in both cases the reason for the asymmetry is built into the modalities themselves.
The moral of the story, of course, is that, except in very specific cases, ethicists should never be talking about what's permissible; they should be talking about what's impermissible and what makes it so, precisely because of this sharp difference in strength between the two.