Let us fix j<q_{R_A} which defines a highest and lowest weight \mu_{top},\mu_{bot}\in \Delta_{R_A}. First, consider the highest weight component. In general, this belongs to a spin-s_I representation of SU(2)_I. The fermions \psi^a_{R_{A,s_I}} for \mu_a \in \Delta^{(j)}_{R_{A,s_I}} are of the form
\begin{aligned}\begin{split}
\psi_{R_{A,s_I}}^a=\sum_b {\cal D}^{(s_I)}_{s_I-a,s_I-b}(\theta,\phi)\left(\begin{array}{c}
f_{b}(r)\,e^{i \frac{\phi}{2}}{\cal D}^{(n)}_{-m-s_I+b,Q_{s_I}+\frac{1}{2}}(\theta,\phi)\\
g_b(r)\,e^{-i\frac{\phi}{2}}{\cal D}^{(n)}_{-m-s_I+b,Q_{s_I}-\frac{1}{2}}(\theta,\phi)
\end{array}\right)~.
\end{split}\end{aligned}
Since the \widetilde{\cal D}_\infty eigenvalues of such a solution are \lambda_\pm^{(n)}=\pm \sqrt{(n+1/2)^2-Q_{s_I}^2}, it follows that there exists a spin down zero-eigenvector of \widetilde{\cal D}_\infty with n=Q_{s_I}-\frac{1}{2} iff Q_{s_I}-\frac{1}{2}\geq0 and it is the unique zero-eigenvector belonging to this {\vec{I}}-representation. Conversely, if Q_{s_I}\leq 0, then there is no spin down zero-eigenvector from the irreducible {\vec{I}} representation.
Thus, let us assume that Q_{s_I}>0. This condition combined with q_{R_A}>j,s_I implies that
\begin{aligned}\begin{split}
Q_{s_I}-\frac{1}{2}+s_I\geq j\geq Q_{s_I}-\frac{1}{2}-s_I\geq0~.
\end{split}\end{aligned}
Now, let us pick n=Q_{s_I}-\frac{1}{2}\geq0. We now must show that the solutions constructed above have a non-trivial total spin-j solution which requires showing that [j]\in [n]\otimes [s_I].
By using the standard multiplication of Wigner D-functions, we can expand the product of Wigner D-functions in (Proof) as representations with total spin:
\begin{aligned}\begin{split}
\left[Q_{s_I}-\frac{1}{2}\right]\otimes \left[s_I\right]=\left[Q_{s_I}-\frac{1}{2}-s_I\right]\oplus \left[Q_{s_I}+\frac{1}{2}-s_I\right]\oplus...\oplus\left[Q_{s_I}-\frac{1}{2}+s_I\right]~,
\end{split}\end{aligned}
where [p] is the representation of total spin-p. It then follows from (?) that [j]\in \left[Q_{s_I}-1/2\right]\otimes \left[s_I\right] and that there is a single spin down zero-eigenvector of \widetilde{\cal D}_\infty in the irreducible {\vec{I}} component containing \mu_{top} when \langle \mu_{top},\gamma_m\rangle>0. The explicit solution is given by
\begin{aligned}\begin{split}
f_a=0\quad, \quad g_a=\begin{cases}
\left\langle j,Q_{s_I}+s_I-a-\frac{1}{2}\big{|}Q_{s_I}-\frac{1}{2},Q_{s_I}-\frac{1}{2};s_I,s_I-a\right\rangle&\mu_a\in s_I\\
0&else
\end{cases}
\end{split}\end{aligned}
where we have chosen a basis \{\mu_a\}_{a=0}^{2s_I} of \Delta_{R_{A,s_I}}
Now, since the eigenvectors of \widetilde{\cal D}_\infty decompose with respect to irreducible {\vec{I}}-representations, it follows that the unique spin down zero-eigenvector of \psi_{R_A} with total spin-j of \widetilde{\cal D}_\infty is the one constructed above containing \mu_{top}\in \Delta_{R_A}. The reason is that any spin-j eigenvector F_0^{(j)} of \widetilde{\cal D}_\infty that belongs to a different {\vec{I}}-irreducible component of spin-s_I' (i.e. the spin-s_I' representation does not contain v_{\mu_{top}}) will not be spin polarized as we have shown in Section ? by explicit construction. This follows from the fact that [\gamma_m,{\vec{I}}]=0 which, when restricted to an irreducible SU(2)_T representation implies Q_{s_I}\neq Q_{s_I'} even when s_I=s_I'.
Then, since the gauge transformation \omega that maps the abelian, asymptotic gauge to the canonical gauge commutes with \vec{S}, any such F_0^{(j)} will necessarily correspond to a solution that is not polarized and hence with quantum number n in the asymptotic gauge such that
\begin{aligned}\begin{split}
n\neq \begin{cases}
Q_{s_I'}-\frac{1}{2} & Q_{s_I'}>0\\
-Q_{s_I'}+\frac{1}{2}&Q_{s_I'}<0
\end{cases}
\end{split}\end{aligned}
Using the fact that \lambda^{(n)}_\pm=\pm \sqrt{(n+1/2)^2-Q_{s_I'}^2}, we see that any such F_0^{(j)} cannot be a spin-j zero-eigenvector of {\cal D}_\infty and that the only spin-j zero-energy solution is the one constructed above.
Therefore, we see that there is only a single spin down zero-eigenvector of \widetilde{\cal D}_\infty with total spin-j iff \langle \mu_{top},T_3-I_3\rangle>0 coming from the irreducible {\vec{I}} component containing \mu_{top} as constructed above.
A nearly identical argument shows that there only exists a spin up zero-eigenvector of \widetilde{\cal D}_\infty when \langle \mu_{bot},T_3-I_3\rangle<0.