I read in serveral books about the definition of a spherical distribution; $x \in \mathbb{R^n}$ has a spherical distribution if and only if $x \stackrel{d}{=}Ox$ for any orthogonal, $n$-dimensional matrix $O$. But now I try to understand that; Let's take the "Von Mises–Fisher distribution"; If you don't know it, it's absolutely no problem, it's just a distribution on the $n$-dimensional unit sphere.
https://en.wikipedia.org/wiki/Von_Mises%E2%80%93Fisher_distribution
Now considering that it's distributed on the sphere, it would make sense that the distribution is a spherical distribution; But if you check the picture from the Wikipedia article and see the red points on it, then you see that the distribution can be very concentrated on a point on the sphere; But if $Qx$ now has the same distribution as $x$, then it would be invariant regarding rotations but it seems unlogical that the distribution stays the same, since the concentration point $\mu$ would change.
So where is my mistake?