No, it does not. The CM is the point $\mathbf{r}_0$ that minimizes
$$\int d\mathbf{r} \, \rho(\mathbf{r}) |\mathbf{r} - \mathbf{r}_0|^2.$$
The point that minimizes the GPE instead maximizes
$$\int d\mathbf{r} \, \rho(\mathbf{r}) |\mathbf{r} - \mathbf{r}_0|^{-1}.$$
Mathematically, the CM minimizes the second moment of the distribution, while the GPE is the point that maximizes the first inverse moment. One can similarly define points that extremize any desired moment $\mu_n$. For example,
- the point that minimizes $\mu_1$ is the generalization of the median to 3D space
- the point that minimizes $\mu_\infty$ is the center of the smallest sphere that contains the mass
- the point that maximizes $\mu_{-\infty}$ is the exact point where $\rho$ is largest
These are all generally different points. In general, as $n$ increases, $\mu_n$ begins to count all masses equally, while for lower $n$ the densest regions have an oversized contribution. There is a bit of a discontinuity at $\mu_0$ though, so I'm not sure how to directly compare $\mu_2$ and $\mu_{-1}$, but there's no reason for them to coincide in general. But at the least, $\mu_{-1}$ should lie "between", say, $\mu_{-2}$ and $\mu_{-\epsilon}$.
If you want bounds on the inverse moments, or relations between them, there are presumably lots of papers on the subject, though at that point this is just a math question.