"In maths but not for real" is just the opposite of a successful model. Either they do hybridize in the observations and the corresponding maths are useful, or both are wrong.
Or in intermediate cases, which is the case for hybridization, the model is known to be wrong, is outdated, but is still useful for it simplicity. Then the math operations shall please give at least some results that resemble some observations.
I wouldn't put that "orbitals are mathematical objects only". Wave functions are observable and observed, precisely in the case of orbitals. They define, when they're written as functions of xyzt, from which locations the particle acts simultaneously and with which amplitude and phase. Few things could be less real, even if the observation isn't always trivial, and even if we have no simple mathematical expression for them most often.
And more generally, I hate commenting wording details about QM. That's the wrong approach.