Money has diminishing returns because if getting 1 util takes 1 dollar, getting the second util is likely to take more than a dollar. For example, if you want want save 5,000,000 lives, it might be fairly cheap to spend money saving the first 1,000,000 lives, but after a while you'll have to start dumping more and more money into R&D. The point is that money is not directly what you care about, which is why money has diminishing marginal utility.

However, if lives are precisely what you care about, then you should discount large numbers of lives only if your utility function exhibits diminishing marginal utility for lives. Does your utility function exhibit this? How do you know? Sure, in econ, most if not all goods exhibit this behavior, however, economists are trying to describe behavior. They don't care about things like scope insensitivity because it's just irrelevant to their purpose - they're trying to predict what you're going to do, not determine what you value. The two are related, but not the same (unless you assume away all mistakes).