One strategy David Boonin makes use of repeatedly in The Problem of Punishment is to show that putative justifications for legal punishment justify more than legal punishment: they justify punishing the innocent.
Boonin deploys this strategy against the act-utilitarian defense of legal punishment, which holds (as you might expect) that legal punishment is justified because punishing all and only offenders brings about more beneficial consequences, net, than any alternative approach to punishment. For punishing the innocent, Boonin imagines a scenario in which a riot will occur unless an innocent person, who everybody except the state believes to be guilty, is imprisoned. Punishment will bring about more good consequences than no punishment here, but it seems wrong to punish. At the very least, act-utilitarianism fails to show that the line between offenders and non-offenders is morally relevant.
Boonin also deploys this strategy against the rights-forfeiture defense of legal punishment, which holds that, in offending, offenders give up their moral rights against a certain range of punishments. For punishing the innocent, Boonin imagines a scenario in which a husband scrupulously studies the laws against spousal abuse and then, short of breaking the law, does everything he can to abuse his spouse. It is hard to see why this person fails to forfeit his moral right against punishment if an offender — just on the other side of the legal line — would not.
The punishing the innocent objection is powerful because it simultaneously (1) shows that a putative justification for punishment does not, in fact, entail that punishment (as opposed to what one might term “punishment+”) is justified and (2) points out an implication of a putative justification for punishment that runs contrary to most people’s moral intuitions, and thus spells its moral doom. However, a (at least rhetorical*) problem with Boonin’s emphasis on the punishing the innocent objection is that it both relies on jerry-rigged, eccentric cases where punishing the innocent seems justified and leaves our intuitions about the moral permissibility of intentionally inflicting authorized reprobative retributive harm on a core class of offenders undisturbed. We are therefore motivated to find a fix instead of to give up our commitment to the moral permissibility of punishment.
* I say “at least” because it could well be that the rational reaction to an argument that attacks a strongly held position using jerry-rigged, peripheral cases while leaving core cases untouched is to continue to hold the position while looking for a fix.