There is a straightforward reason why anyone interested in social progress should insist that the justice system be basically retributive in nature: the retributive theory of punishment is the only theory of punishment that intrinsically depends on such concepts as justice, guilt, and innocence.
There are, of course, ways to improve a retributive system, but all involve keeping a basic retributive core. For instance, one can moderate a strict retributive system with consequentialist concerns; one can adapt the particular modes of punishment so that they are most fit to deter others from the crime; one can add to the system a concern for rehabilitation and reparation; and so forth. If these other elements (social consequences, deterrence, rehabilitation, reparation, etc.) are taken as rivals to retribution, however, then the only theory that keeps the state within sharp bounds is the retributive. Because retribution theory is intrinsically justice-oriented, it sharply limits what a state can do: the state's powers of punishment are to be applied to people who are guilty of crimes, end of story. One can have more or less strict versions of this, but they all impose the same basic justice-based limitation. The rivals, however, unless they are themselves limited by retributive principles, provide no barrier to punishing those innocent of any crime (because of the good social consequences, or the deterrent effects, or an expansive notion of rehabilitation and reparation).
This is only the sketch of an argument, but I think it can be developed more rigorously so as to meet the objections of partisans of other theories. The problems with a purely rehabilitative system, for instance, have been discussed at length by G. K. Chesterton and C. S. Lewis, along lines at least roughly similar to the one I have suggested. Nothing says a justice system has to be confined to retribution; but it certainly needs to be retributive at heart.