What is wrong is the use of pair style hybrid. It is wrong it two ways: a) it doesn’t make sense in terms of the physics. Both of the potentials are manybody potentials and have different ways to include the manybody interactions. You cannot just say that you want one for some pairs and another for some other pairs. That is not how manybody interactions work, and there is no consistent way to enter these. Even if there would be a way, the result would be bogus.
And b) I don’t think that COMB (or COMB3) for your LAMMPS version is compatible with a hybrid potential at all.
The only way to model your system consistently would be to use COMB3 for all interactions.