Is It Good When More People Are Endangered by Trolleys?
Moderate deontology thinks so
Moderate deontology holds that one should not kill one person to prevent slightly bad outcomes, but they should to prevent sufficiently disastrous outcomes. Thus, you shouldn’t push a person off of a bridge to prevent five deaths, but you should to prevent a billion deaths. Moderate deontology, being non-consequentialist, regards the world where one person dies from being pushed off the bridge as better than the world where five people have died from being pushed off the bridge, but thinks that it is impermissible bring about that better state of affairs because of constraints.
But this produces a very strange result. Consider two possible worlds. In both cases, a perfectly moral person is on top of a bridge and can push a man to stop a train. In one of the worlds, if they don’t push the person, a billion people will die. In the other world, five will die. Moderate deontology has to hold, as a result of these commitments, that the world where a billion are endangered is better, because it will involve the person pushing the man off the bridge.
Deontology is a framework for evaluating actions. Utilitarianism evaluates actions by collapsing the distinction between actions and states of affairs, that is, we determine the goodness of an action by comparing the states of affairs that arise from taking or not taking the action. Deontology cannot be used to evaluate states of affairs, because states of affairs are not actions and do not take actions. So, deontology doesn't render verdicts on states of affairs, and therefore doesn't render the wrong verdict on states of affairs.
If the argument is that taking actions that are deontologically sanctioned brings about bad states of affairs, then you don't need to say anything more than: "Deontology positively evaluates actions that bring about states of affairs which, using a separate moral framework, I assess as bad," which you do in the first premise.
But that grants that you're evaluating states of affairs, and if you are evaluating states of affairs, then you are using (at least) something other than deontology to do so, and in your case, you are using utilitarianism.