$A_{\infty}$-structure on closed manifold

Yes:

As has been pointed out, admitting the structure of a connected $A_\infty$-space is the same thing as being homotopy equivalent to a connected loop space.

The Hilton-Roitberg criminal, mentioned by Vladimir Dotsenko, was the first example of a closed manifold homotopy equivalent to a loop space, but not to a Lie group. The space has the rational homotopy type of the compact Lie group Sp(2).

In fact there exists "rational criminals", i.e., finite loop spaces that are homotopy equivalent to closed smooth manifolds (as always, by the Broto-Kitchloo-Notbohm-Pedersen theorem), but not even rationally homotopy equivalent to a Lie group. However, they only start occurring in rank 66 and onwards.

See this paper that I wrote with Andersen-Bauer-Pedersen

If you require just "not homeomeorphic", then there are very silly examples of all sorts. What you want to ask is "not homotopic", I suppose.

For that, I know some useful references in "Which H-spaces are manifolds? I" by Cappell and Weinberger (http://www.sciencedirect.com/science/article/pii/0040938388900171), e.g. you can look at the manifold appearing in Formula (1.4) in "On Principal S3-Bundles Over Spheres" by Hilton and Roitberg (http://www.jstor.org/stable/1970683).

Yes, any connected A-infinity space with finitely generated homology is homotopy equivalent to a closed smooth connected manifold, by results of Bauer-Kitchloo-Notbohm-Pedersen. Many of those are known not to be homotopy equivalent to Lie groups (see references therein).