Let's consider the consequences of that line with respect to electoral math. If we consider only serious criminals, e.g. murderers, they constitute a negligible proportion of the population and with high probability the number of election outcomes changed by allowing them to vote or not would be none.

By contrast, if you lump in people convicted of things like drug possession, that is enough people to change the outcome of some elections. And in general it's a strong heuristic that if huge numbers of people are committing a particular crime, it's a result of flaws in the law or society rather than flaws in huge numbers of different people.

So the only time disenfranchising felons matters to the outcome is when you get the line wrong, implying that it shouldn't be done because it shouldn't affect the outcome unless it's being done improperly.