I am having trouble with Q. Liu's proof of the classification of quasi-coherent sheaves on an affine scheme. Specifically, I don't understand the first step of his argument (proposition 1.6., section 5.1).
He starts by saying that, by definition of quasi-coherent sheaf one always has an affine open and a sequence $$\mathcal{O}^{J}_V\to\mathcal{O}^{I}_V\xrightarrow{\alpha}\mathcal{F}_{|V} \to 0$$
He does not prove this claim, but this should be easy: by definition of quasi-coherent sheaf one has a generic open $U$ and an exact sequence as above, and then one can restrict to an affine open $V \subset U$ and check the exactness of the sequence on the stalks.
question What I don't understand is what he does next: he claims that considering the $\mathcal{O}_V(V)$-module $\operatorname{Im}\alpha_V$ and $\tilde{\operatorname{Im}\alpha_V}$, one has an exact sequence $$\mathcal{O}^{J}_V\to\mathcal{O}^{I}_V\xrightarrow{}\tilde{\operatorname{Im}\alpha_V}\to 0$$
I do not see why he has that exact sequence.
First of all, by the right exactness one has $\operatorname{Im}\alpha_V \cong \mathcal{F}(V)$, so he might as well call it $\mathcal{F}(V)$.
Next, I think he uses, tacitly, the fact that $\mathcal{O}_V \cong \tilde{\mathcal{O}_V(V)}$, which can be proved by observing that $$\mathcal{O}_V(D(f)) \cong (\mathcal{O}_V)_f \cong \tilde{\mathcal{O}_V(V)}(D(f))$$ and hence they are isomorphic on stalks.
specific question
I don't see how does he get the sequence $$\mathcal{O}^{J}_V(V)\to\mathcal{O}^{I}_V(V)\xrightarrow{} \mathcal{F}(V)\to 0$$
because I know that exactness on the right of a sequence of sheaves does not in general imply that the sequence must be right exact on each open subset. So why can he infer this in that case?
First, note that the exponents here correspond to direct sums, not products. Let $\beta: O_V^{J} \rightarrow O_V^{I}$ map: then by the above exact sequence, $\alpha$ (or $\mathcal{F}_{|V}$) is the cokernel of $\beta$.
Let $M$ be the cokernel of $\beta(V): O_V^J(V) \rightarrow O_V^I(V)$: it’s an $O_V(V)$-module, and the sequence $$O_V^J(V) \overset{\beta(V)}{\rightarrow} O_V^I(V) \rightarrow M \rightarrow 0$$ is exact.
Apply the tilde functor: $O_V^J(V)$ becomes $O_V^J$, $O_V^I(V)$ becomes $O_V^I$, and $\beta(V)$ becomes $\beta$, so that we have an exact sequence
$$O_V^J \overset{\beta}{\rightarrow} O_V^I \rightarrow \tilde{M} \rightarrow 0,$$
and thus $\tilde{M} \cong \mathcal{F}_{|V}$.