I think that the statement is **false** for all ruled surfaces over a curve of strictly positive genus.

In fact, let $C$ be a curve of genus $\geq 1$ and let us consider two distinct points $x, \, y \in C$. Take now the two rank $2$ vector bundles on $C$ defined by $$\mathscr{E}_x = \mathscr{O}_C \oplus \mathscr{O}_C(x), \quad  \mathscr{E}_y = \mathscr{O}_C \oplus \mathscr{O}_C(y).$$

I claim that the two projective bundles over $C$ given by $\mathbb{P}(\mathscr{E}_x)$ and $\mathbb{P}(\mathscr{E}_y)$ are *not* isomorphic, hence the birational class of $C \times \mathbb{P}^1$ contains *uncountably many* distinct minimal models.

The argument is as follows. By contradiction, assume that they are isomorphic. Then there exists a line bundle $\mathscr{L}$ on $C$ such that $\mathscr{E}_x \otimes \mathscr{L}=\mathscr{E}_y$, that is $$\mathscr{L} \oplus \mathscr{L}(x) =  \mathscr{O}_C \oplus \mathscr{O}_C(y).$$

Since by the Krull-Schmidt theorem the decomposition of a vector bundle into indecomposable summands is unique up to permutations of the summands, we must have either $\mathscr{L} = \mathscr{O}_C$ or $\mathscr{L}=\mathscr{O}_C(y).$    

Now, both cases are impossible: the former since $x \neq y$ and the curve $C$ has strictly positive genus, and the latter by degree reasons.