Well you are right about drones but its still a wrong analogy. The system we are talking about is one used to shoot down incoming projectiles. The missile controller should be able discern between a dummy and a real destructive incoming missile. No doubt the system will be fine tuned to distinguish shape/speed/temperature of the incoming projectile and respond accordingly.
There is a point where having a decoy is pointless -- and that point is the descent stage.
Think about this for a while...And bear with me a little on the language front...
A feint is not a ruse. Anyone who engages in combat, whether in a sporting environment or in a war, knows that a feint is not really a feint.
A ruse is a genuine tactic at misdirection while a feint is actually more complex.
A feint must always be readied to turn into a genuine strike. A feint is for when, if an opponent chose to respond, you withdraw and strike from a different direction, but if he chose not to respond, you will turn that feint into a strike and that you must be readied to follow up on that strike. You have an original strike in one direction, but you are prepared to either change direction or use both directions. In other words, a feint is a strike of opportunity.
Since a ruse is not designed to be a strike of opportunity, the ruse can be designed to be expendable and discard when no longer useful. On the other hand, since a feint was designed to be strike of opportunity, that feint must be constructed to be as powerful and effective as the method you originally intended to use.
What does this have to do with ICBM design philosophy ? Everything.
In the missile's ascent stage, a decoy would be a ruse. At this stage, your real missile is your genuine strike while the ruses are designed to be expendable and intends to confuse/mislead any observer. The decoy missiles must have the same flight behaviors as the real strike missile. If anything, the only difference between the real strike and the decoy is the warhead section. The negative side to this is that it is financially prohibitive to design, build, and maintain 9 decoys to mislead the observer from that one real missile.
So the next best place to have and deploy decoys (ruses) is the mid stage. This is where the missile's true target have a certain amount of statistical uncertainty as to where its true target(s) is/are. So you have one missile that will deploy multiple vehicles and only one of them is the real warhead.
So why not have the decoys at the descent stage ?
The better question is 'Why should you ?'
Once a warhead is in the descent stage, ballistic physics reduces most, if not all, amount of statistical uncertainty as to its intended target. Even if the warhead is maneuverable, it does not mean its intended target cannot be deduced. It is like watching a car coming at you while it is changing directions and you are fixed.
So why should you design a decoy at the descent stage ? Why not make it a real missile ? The warhead is already on ballistic certainty -- to the ground. Why not make it a real strike ? If you launch two real missiles, one of them could be a feint, a target for an interceptor. If the interceptor failed, now you have greater assurance that the intended target will be hit. Take artillery, for example. How many artillery rounds are fake ? When you have carry a rifle into combat, how many rounds are fake ? If you are a sub hunter, how many torpedoes and depth charges are fake ?
At the descent, aka 'terminal', stage, just as it is pointless for you to deliver a fake warhead, it is equally pointless for me to try to guess which of the multiple descending vehicles is the real warhead. It is the endgame for both of us. If my interceptor hit your
ONE real warhead out of 2 vehicles, I win. But if my interceptor hit one of two real warheads, you win.
Ultimately, you can design your strike any way you want based on any combat doctrine. But a war is not an experiment.