John Gray writes:

I assume that a moderate deontologist would not allow the murder of one
innocent human being if doing so were necessary to prevent two deaths,
or perhaps even three or four deaths. My question is: what standard
does the moderate deontologist use to determine that something is evil
enough to allow for the performance of an intrinsically wrong act that
will prevent that evil? It seems to me that a moderate deontologist is
someone who is sometimes a deontologist, and sometimes a
consequentialist. Which is to say, a moderate deontologist is sort of a
moral pluralist.

Moderate deontologists (unlike absolutist deontologists) must set a threshold. A high threshold would require, say, the saving of 1,000 lives to justify killing one innocent person. A low threshold would require, say, the saving of 50 lives. Moderate deontology is a distinct theoretical position; it is neither identical nor reducible to either absolutist deontology or consequentialism. Sometimes the moderate deontologist agrees with the absolutist deontologist; sometimes the moderate deontologist agrees with the consequentialist. Sometimes all three agree. But agreement isn't identity. John is right about the final point: moderate deontologists are pluralists. What justifies the killing of the innocent is overall good. As for the "standard" the moderate deontologist uses, there is none. The theory sets out relevant considerations, but provides no decision procedure. The person applying the theory must exercise judgment as to the weight or strength of these considerations. See, for example, W. D. Ross, The Right and the Good (1930), esp. chap. II ("What Makes Right Acts Right?").