I think I managed to prove that, if a finite group $G$ has an automorphism $\varphi$ of order $2$ which doesn't fix any nontrivial elements, then:
- $G$ has odd order;
- $G$ has odd conjugacy classes, and each of them is of odd size;
- $\varphi$ acts on the set of the conjugacy classes of $G$ likewise it acts on $G$ itself, namely it doesn't fix any nontrivial conjugacy class.
This is not surprising, since such a group turns out to be Abelian (see e.g. here and here). I'd expect 1-2-3 to lead to an alternative proof of $G$'s Abelianiness, by obtaining from them that all the conjugacy classes must be singletons. This is where I'm stuck.