In Sakurai's and Napolitano's book "Modern quantum mechanics" there's a nice proof of the theorem. This can be found also almost identical on Wikipedia's Wigner–Eckart theorem - Proof.
The thing that is bugging me about this proof is this part
We therefore have two sets of linear homogeneous equations \begin{align}\sum_c a_{bc} x_c &= 0 & \sum_c a_{bc} y_c &= 0\end{align} one for the Clebsch–Gordan coefficients $x_c$ and one for the matrix elements $y_c$. It is not possible to exactly solve for $x_c$. We can only say that the ratios are equal, that is
$$\frac{x_c}{x_d} = \frac{y_c}{y_d} \quad\text{or}\quad x_c=ky_c.$$
While the rest is clear, I don't get how we conclude the proportionality. In general, this is clearly not true. This is true only if the kernel of the matrix $a_{bc}$ is one-dimensional. So there must be some other property of the coefficients $a_{bc}$ that allows us to conclude that the kernel is one-dimensional, and thus the proportionality of different $x_c$ satisfying the equation.
 
     
     
    