This is not the result I was expecting, I though I would get what we call an inverted color. I get the same issue with everything: gradient, texture (sRGB and linear)… And I also get the same result when I use a Substract node instead the OneMinus.
It looks like there is a gamma applied somewhere, but why?
Would it be possible to have some explanations please? I am a bit lost (and struck) there.

What you are showing in the pictures above is giving you an inverted (1/#) output from your LERP. But for the specific example above you are using a node which generates a curve based on exponential input (Fresnel) and the LERP is trying to work within a 0 to 1 space so going over a power (exponent) of 1 will cause the LERP to try to push the values further into the B input. For your setup is that while on the center of the sphere has a value of 0 the edge may have a value of 1.75, so taking the inverse of those numbers would give you 1 in the center and 0.5 (ish) on the edge. which may not be perceived as a “correct” inverse in a clamp from 0 to 1 linear space. If you run the test using a perfect gradient like a masked Texture Coordinate you can see that the One Minus is working as intended.