Although angles are dimensionless for the purposes of dimensional analysis, a choice of "angular unit" must be made in order for an angle to be considered as a real number, a necessary step in constructing trigonometric functions which we can use calculus on. (In fact when we do things rigorously it is quite a pain to properly axiomatise geometry, and the trig functions are usually defined by power series instead.)
Regarding your last paragraph in particular, what are we inputting 90 "degrees" into? The function f(x):=x defined on some domain D takes an element of the set D and spits out the same element. When we are doing calculus, we would like D to be a subset of real numbers, some Euclidean space or a manifold etc. This means that the choice of units for whatever physical/geometric object we are measuring occurs before the function gets introduced, so if we want to talk about the function that maps an angular measurement in degrees to an angular measurement in radians, this is usually done with the function f(x)=pi*x/180, not pi*x with some condition that the y-axis and x-axis have different scales.
All of this is mostly a matter of convention and mathematical practice, rather than a matter of any of our statements being "incorrect" so I don't really want to drag out this discussion further, but that's about all I have to say about it.