Not every system is IC.
It's possible that in biology, no systems are IC in the sense that Behe postulates - in other words, that they cannot have been arrived at via natural processes.
Not every item is part of a system.
I'll grant that, but you can't shift the goalposts and define every system as an IC system purely on the basis that if a part can be removed and the system doesn't fail, it obviously wasn't part of the system to begin with. "It doesn't fail" isn't the same as "that part served no purpose
and never did".
Let's see if I can use an illustrative example, which I'll hope I don't screw up, though given the lateness of the hour I can't make promises.
Imagine a simulation with six possible components - A, B, C, A', B' and C'.
In this simulated environment, for a system to be functional, C requires A, and A' and C' require B', and at least one of each letter must be present in any system of three or more parts. Further, no system of fewer than three parts may have a "dashed" letter.
Let's say you have a system of three parts, A, B and C. (Which, according to the simulation, can be arrived at via simple addition of the parts in the sequence shown.)
Later there's a mutation that alters B to B'. Now we have A, B' and C. By the rules above, our system is functional.
Later on there's an additive mutation that introduces part A'. Now we have A, A', B' and C. Still functional.
Finally, there's a mutation that alters C to C'. The alteration to C' renders A redundant: C requires A, but nothing else does. We still have a functional system.
Note, this doesn't necessarily mean A serves no purpose; only that it is not necessary for the system to function. Note also that we've arrived here entirely via the simplistic rules of the simulation laid down at the beginning of the exercise.
By your argument, what
you now see is an IC system A' B' C', and an extraneous part A that "isn't a part of the IC system".
You would dyke out part A, look at A' B' C' and declare that it's IC. Am I right?
Further - and in response to your earlier confusion about substraction - let's remove A and look at the system A' B' C', which - by your reasoning - is IC. By your (and Behe's) argument, A' B' C' could not possibly be arrived at via natural processes - because removing any single part would render the system non-functional.
By excluding A from consideration (even though it's there), and ignoring all the rules by which we could have arrived at A' B' C' via successive "mutations" of A B C, you've basically excluded even from consideration the means by which, in this case, A' B' C' was actually arrived at.
Can you see the problem with the concept of IC yet?