Just in case there is someone who wants to think about it in another way. (This is how I thought about it.)
The situation is the following. Let $A$ be a ring, $\mathbb{P}^n_A= \text{Proj}(S)$ with $S=A[x_0,\dots,x_n]=\bigoplus_{d\geq 0}S_d$ (standard grading) be the $n$-th projective space over $A$ and let us for short denote by $\mathscr{O}=\mathscr{O}_{\mathbb{P}^n_A}$ the structure sheaf of $\mathbb{P}^n_A$.
What is meant by "$\mathscr{O}(1)$ is generated by the global sections $\{x_i\}_i$" is that for every point $P \in \mathbb{P}^n_A$, the stalk $\mathscr{O}(1)_P$ is generated over $\mathscr{O}_P$ by the germs $\{{x_i}_P \}_i$ of the global sections at $P$. But here one has to be carefully (at least at the first attempt): the $x_i$ are not "really" global sections of the twist $\mathscr{O}(1)$. At first we have to clarify how they give rise to global sections of $\mathscr{O}(1)$.
By definition, \begin{align*} \mathscr{O}(1)=S(1)^{\sim}=(\bigoplus_{d\geq 0}S_{d+1})^{\sim}=(S_+)^{\sim}. \end{align*} Then $x_i$ gives a map \begin{align*}
X_i \colon \mathbb{P}^n_A &\longrightarrow \coprod_{P \in \mathbb{P}^n_A} (S_+)_{(P)}, \\ P & \longmapsto \big(\frac{x_i}{1},P\big),
\end{align*} (here note that $\frac{x_i}{1}$ is a degree zero element in the localization $(S_+)_P$, since $x_i$ is a degree zero element in the grading of $S_+$) which is a global section of $\mathscr{O}(1)$.
Now using the natural isomorphisms \begin{align*}
(\mathscr{O}(1))_P \cong (S_+)_{(P)} \;\;\; \text{and} \;\;\; \mathscr{O}_P \cong S_{(P)},
\end{align*}
we see that it is enough to show that the elements $\frac{x_i}{1} \in (S_+)_{(P)}$ (corresponding to the germs of the $X_i$ at $P$) generate $(S_+)_{(P)}$ over $S_{(P)}$.
But $\{ x_i \}_i$ generates $S_+$ over $S$, hence we get that $\{ \frac{x_i}{1} \}_i$ generates $(S_+)_{(P)}$ over $S_{(P)}$ as desired.