In your use case.
If the R channel is 0 you get what?
What is 1 * 0?
Lets see if you are able to put 2 and 2 together…
Once you realize you are wrong,
Have a look at HLSL code for the proper way you should do the saturation bump.
Just because you can simplify something, doesn’t mean you should, unless you have some kind of certainty that a result will avaliable;
Using a random texture (which may not even load btw) doesn’t give you any kind of assurances for that R channel.
Additionally to that, you are clamping the result - which isnt wrong, but what is 1×1 ?
And what is .1 x 1 ?
Or the commutative property of the multiplication?
Even just logically without knowing math:
“Hey system, do Y * X, and then just dump all the stuff thats above 1.”
Is that logical?
No… Maybe Normalize would be (if you also had something that factually took the values out of range):
“Hey system, re-adjust everything so that the highest value you got is always 1 and everything else is scaled down appropriately”.
Anyway, the point is: the code is wrong.
The HLSL is not.
Implement a custom node with that directly if you don’t care to learn.
Then your solution could work - if you overlook perfance concerns entierly.