If I delete them, the cone turns into a wall or a pyramid if i add the necessary abs()'. There is no way you could get a cone without using at least the x2,z2 nodes.

This is the way how I understand it works:

X to scalar: (x,y,z) -> x (the function assigns to a vector its x-component [and doesn't remember that this has once been an x-value])

Y to scalar: (x,y,z) -> y

...

A scalar, that handled xyz seperately would be a vector. Why would we need the concept of a scalar if it was just a special form of vector?