Using one principle (nukes are expensive) isn't the same thing as deducing something from first principles. In particular, it's not enough, especially given hypersonic missile technology, to just take an ICBM, reprogram it, and voila you have an ABM. Because the cost of even a single warhead off an MIRV ICBM getting through is so devastating, the cost of a ABM system includes a huge number of missiles to increase the chance of catching the incoming attack. And then, once the ABM missile reaches the incoming ICBM. How does it neutralize the threat? What technology do we have that causes a lot of destruction? Oh that's right, nukes. If the ABM system uses nukes to neutralize the incoming threat, then it's very much not clear that the singular principle (nukes are expensive) doesn't also apply to the ABM system.
Some subjects, like computer science, which is still a very young area of study are very open to learnings from first principles. Others, like the geopolitics of nuclear deterrence, are deeper than any one person can even begin to understand.
Current interceptors do not use nuclear warheads. This makes it both cheaper and more expensive. It requires a more complicated hit to kill system than merely getting the nuclear warhead close enough. However, the system as a whole ends up cheaper. You don't need to get National Command Authority to authorize a launch anymore, because you're not using nukes. You can get that delegated down to whichever officer is in charge of the system that day.
The objective of National Missile Defense isn't to stop all the warheads, it's to make the attacker's job harder. There's a bunch of ways it does that. One is that the defender gets to choose what it stops. So they can stop all the warheads going to one target. But the attacker doesn't know ahead of time, so they need to launch enough warheads at each target to destroy it regardless of whether it's actually defended or not. So 1550 warheads no longer hit 750 targets, but rather they hit 75 targets. It really sucks to be one of those 75 targets, because most of them get absolutely destroyed, but the other 90% of potential targets are much better off, because they didn't get hit at all.
Another way is that ballistic missile accuracy depends on how far away from the target it gets separated from the missile bus that has the guidance system. If they separate too late, the interceptor will hit the bus before separation. Which kills all the warheads. So they have to move it back, which makes it less accurate, which means it's more likely the warhead will miss the target.
A third way is that they may decide they need to carry penetration aids. But there's a bunch of really fancy tricks radars and IR sensors can do to see which is a real thing, and which isn't, which amount to: the penetration aid which is best at mimicking a warhead ends up being the same size and shape as a real warhead. So they might as well just use a real warhead.
Some subjects, like computer science, which is still a very young area of study are very open to learnings from first principles. Others, like the geopolitics of nuclear deterrence, are deeper than any one person can even begin to understand.