This one is quite deeply rooted. We can see this by looking at two easily accessible and influential sources: Alexander Moseley's Just War Theory article at the IEP and Brian Orend's War article at the SEP. I think there are problems with both, but of these, Moseley has a more historically correct understanding of proportionality and why it is a criterion; but Orend summarizes what seems to be the most common -- and not very critically examined -- idea about what it is.
The final guide of jus ad bellum is that the desired end should be proportional to the means used. This principle overlaps into the moral guidelines of how a war should be fought, namely the principles of jus In bello. With regards to just cause, a policy of war requires a goal, and that goal must be proportional to the other principles of just cause. Whilst this commonly entails the minimizing of war’s destruction, it can also invoke general balance of power considerations.
This is right (although it's a little odd to talk about the end being proportional to the means rather than vice versa). The 'just war criteria' weren't just put together at random; they arise from considerations of the nature of action itself. All deliberate actions have ends; deliberate actions involve chosen means to those ends. Assuming that the ends themselves are rational and appropriate, to be rational and appropriate, the means must be proportionate to the ends. The proportionality here is exactly the same as we have in the phrase 'sense of proportion', although, of course, war is always a somewhat greater test of our sense of proportion than most other situations. The most important idea here is that the means are to be proportioned to the end. And since we're interested in justice in war, the end is the just cause. Thus we are not interested in abstract calculations here, although calculation may be necessary; the question raised by proportionality is this: "Is the party in question fighting in a way consistent with, and appropriate to, a just cause?"
But we get a somewhat different idea elsewhere. Here's Orend:
A state must, prior to initiating a war, weigh the universal goods expected to result from it, such as securing the just cause, against the universal evils expected to result, notably casualties. Only if the benefits are proportional to, or “worth”, the costs may the war action proceed. (The universal must be stressed, since often in war states only tally their own expected benefits and costs, radically discounting those accruing to the enemy and to any innocent third parties.)
This, although it is a very common idea, is a deviation from the historical concept of proportionality. Note the key point here: the historical idea is about means; this idea makes proportionality about consequences. If you're interested in means, you weigh consequences; but this is hardly all you do.
If I am deciding how to celebrate someone's birthday, I will certainly assess advantages and disadvantages of different ways of doing it, but I will do so only in a relatively limited way, a way limited by my primary concern. My primary concern will be to find something fun for the birthday person that is practicable on the resources and makes sense for a birthday. That is proportionality as traditionally understood: I am looking for a means proportionate to celebration of a birthday. It is no different in war: we are looking for means proportionate to our just cause. We will surely weight benefits and harms, but relevant benefit and harm is determined by the just cause. It is a very different thing from the greatest universal benefit with least universal harm, in which just cause is simply one thing considered and the weighing is unconditional.
As nearly as I can tell, based on some very quick and loose searching, this divergence in positions arises in the late 1960s, when proportionality starts being treated in something like this new way. This is perhaps not surprising; what would be surprising is if the way proportionality is often formulated today preceded the rise of consequentialism, since it is structured like a standard consequentialist rule. There is also something about it that seems to me to require something like a post-WWII political situation; a rule better suited to favoring populous, powerful nations with a rich array of military options would be difficult to think up.
People can use which version they please, but it needs to be recognized that there are two different things floating around under exactly the same name under exactly the same conditions.