Let $A$ be an integral domain and $M$, $N$ be finitely generated $A$-modules. I know from this topic that one cannot expect $\hom_A(M,N)_P \cong \hom_{A_P}(M_P,N_P)$ to be true in the general case (although I lack the background to fully grasp the given counterexample), but
what if I consider the localization of the dual module of $M$, i.e. $N = A$?
For my purposes, I may even assume that $M$ is locally free of rank $1$. If I didn't miscalculate, the natural map should be injective, but I fail to prove surjectivity. (A more abstract proof would be great, too!)