First off, Greiner's Figure $21.9$ is simply wrong. It implies that
$$\vec r=\langle-r\sin\theta\sin\phi,r\sin\theta\cos\phi,r\cos\theta\rangle$$
So
$$\begin{align}d\vec r&=\langle-\sin\theta\sin\phi,\sin\theta\cos\phi,\cos\theta\rangle dr\\
&\quad+r\langle-\cos\theta\sin\phi,\cos\theta\cos\phi,-\sin\theta\rangle d\theta\\
&\quad+r\sin\theta\langle-\cos\phi,-\sin\phi,0\rangle d\phi\end{align}$$
From which we may read off
$$\begin{align}\hat e_r&=\langle-\sin\theta\sin\phi,\sin\theta\cos\phi,\cos\theta\rangle\\
\hat e_{\theta}&=\langle-\cos\theta\sin\phi,\cos\theta\cos\phi,-\sin\theta\rangle\\
\hat e_{\phi}&=\langle-\cos\phi,-\sin\phi,0\rangle\end{align}$$
And then solve to obtain
$$\begin{align}\hat e_x&=-\sin\theta\sin\phi\hat e_r-\cos\theta\sin\phi\hat e_{\theta}-\cos\phi\hat e_{\phi}\\
\hat e_y&=\sin\theta\cos\phi\hat e_r+\cos\theta\cos\phi\hat e_{\theta}-\sin\phi\hat e_{\phi}\\
\hat e_z&=\cos\theta\hat e_r-\sin\theta\hat e_{\theta}\end{align}$$
And that's miles away from what Greiner has. The figure Greiner is actually working from is as you say http://plaza.obu.edu/corneliusk/mp/suv.pdf . From there
$$\vec r=\langle r\sin\theta\cos\phi,r\sin\theta\sin\phi,r\cos\theta\rangle$$
And now we get
$$\begin{align}d\vec r&=\langle\sin\theta\cos\phi,\sin\theta\sin\phi,\cos\theta\rangle dr\\
&\quad+r\langle\cos\theta\cos\phi,\cos\theta\sin\phi,-\sin\theta\rangle d\theta\\
&\quad+r\sin\theta\langle-\sin\phi,\cos\phi,0\rangle d\phi\end{align}$$
And then
$$\begin{align}\hat e_r&=\langle\sin\theta\cos\phi,\sin\theta\sin\phi,\cos\theta\rangle\\
\hat e_{\theta}&=\langle\cos\theta\cos\phi,\cos\theta\sin\phi,-\sin\theta\rangle\\
\hat e_{\phi}&=\langle-\sin\phi,\cos\phi,0\rangle\end{align}$$
And solve to get
$$\begin{align}\hat e_x&=\sin\theta\cos\phi\hat e_r+\cos\theta\cos\phi\hat e_{\theta}-\sin\phi\hat e_{\phi}\\
\hat e_y&=\sin\theta\sin\phi\hat e_r+\cos\theta\sin\phi\hat e_{\theta}+\cos\phi\hat e_{\phi}\\
\hat e_z&=\cos\theta\hat e_r-\sin\theta\hat e_{\theta}\end{align}$$
You see the difference now, right? Greiner has the trig functions of the wrong argument multiplying $\hat e_{\phi}$, while the *.pdf you linked to has the right stuff. Also we can work out that
$$\begin{align}\hat e_r\times\hat e_r&=\vec0\\
\hat e_{\theta}\times\hat e_r&=-\hat e_{\phi}\\
\hat e_{\phi}\times\hat e_r&=\hat e_{\theta}\end{align}$$
And this would lead us to conclude that the overall sign of $\vec H$ in Equation $21.33$ is inconsistent with the previous equation. I think that may be due to forgetting a convention that the charge of the electron is $-e$ earlier on, but then copying the formula into the final result. A very sloppy derivation.
Again, $\vec r=r\hat e_r$ is the vector from the source point to the field point so $\hat e_r$ is the one we need. Where would $\hat e_{\theta}$ or $\hat e_{\phi}$ come in?