This proposition on page 370 is to prove any ruled surface over a nonsingular curve $C$ is $\bf{P}(\mathscr E)$, the projective space bundle of a locally free sheaf $\mathscr E$ of rank $2$ on $C$ and vice versa. In the direction that for a given rule surface $\pi:\, X \to C$, $X \cong \bf P(\mathscr E)$ over $C$, the proof uses the fact that $X$ admits a section $\sigma$. So define the divisor $D = \sigma(C)$ on $X$ and define $\mathscr E = \pi_*(\mathscr L(D))$, where $\mathscr L (D)$ is the sheaf on $X$ corresponding to the divisor $D$. Then $\mathscr E $ is locally free of rank $2$ on $C$ from Grauert's theorem and clearly there is a natural map $$\pi^*(\mathscr E) \to \pi^*(\pi_*(\mathscr L (D))) \to \mathscr L (D).$$ But then there are some parts I do not fully understand (here both $X$ and $C$ are defined over some algebraically closed field):
To prove this map $\pi^*(\mathscr E) \to \mathscr L(D)$ is surjective, it says ''by Nakayama's lemma, it is enough to check this on any fiber $X_y$''. Here $y$ is any point on $C$. But I can not see how to use Nakayama's lemma in this case to prove the surjectivity provided the surjectivity is proved on any fiber $X_y$.
After proving the surjectivity of the above morphism, we can say there is a morphism $g:\, X \to \bf P(\mathscr E)$ over $C$ with $\mathscr L(D) \cong g^*(\mathcal O_{\bf{P}(\mathscr E)}(1))$. I see $g$ is an isomorphism on any fibre because basically in this case $g$ induces a morphism $g_y:\, \bf P^1 \to \bf P^1$ over any point $y$ in $C$ and since $\mathscr L(D)$ is very ample on any fiber, $g_y$ is a closed embedding hence an isomorphism. But I can not see why $g$ itself is an isomorphism. Maybe it is trivial.
Thank you for anyone who can give me clues or details.