Let $X = [P/G]$ be a smooth finite type separated DM-stack over $\mathbb C$ given as the quotient of a smooth projective scheme $P$ by the action of a smooth (finite type separated) reductive group scheme $G$.

At atlas of $X$ is an etale morphism $U\to X$ with $U$ an **algebraic space**.

I'm interested in the converse of the following statement:

**Thm.** If the coarse moduli space of $X$ is a scheme (i.e., not just an algebraic space) then every atlas $U$ of $X$ is a scheme.

*Proof.* The map from an atlas $U$ to the coarse moduli space $X_c$ of $X$ is quasi-finite separated. Therefore $U$ is a scheme  by Knutson Cor. II.6.16., p. 138 QED

So the converse would read as follows:

**Q.** Suppose that every atlas $U$ of $X$ is a scheme. Is the coarse moduli space of $X$  a scheme?

I expect the answer to be negative, but can't find a good example. Note that a counterexample can not be an algebraic space, as an algebraic space with the property that every atlas is a scheme is itself a scheme (the identity morphism being an atlas).

Edit: I have an application in mind of the above question, and in the context of my application the stack $X$ is even generically a scheme. Not sure if that helps...