If you can't draw such a distinction, then your argument proves too much. This is so because your points re:uncertainty of guilt and intractability of punishment don't just apply to capital cases, but to all convictions and punishments, which are entered on the same standard of proof beyond a reasonable doubt. If you believe it's morally unconscionable to convict and sentence with less than 100% certainty, and to impose punishments you can't back out of, then how can any conviction achieved under this imperfect standard ever be considered moral?
Maybe your answer is that they can't be considered moral and we should limit convictions/sentences to cases where we have 100% certainty. But assuming you want some kind of functional criminal justice system for the punishment and incapacitation (oh, and the fictional "rehabilitation") of criminals, then I think you must accept the imperfect system, and distinguish the death penalty from lesser sentences.
This is a distinction you can sensibly draw. My point is just that it's logically important to explicitly do so. (Sorry if this is a tedious, semantic point!)
So how would you draw the distinction? If your concern is with avoiding very bad decisions that are intractable, would you draw that moral distinction to include not only the death penalty but also extremely long (eg, 30+ year) prison sentences?