Okay then. In below, we fix $u$ an eigenfunction of $L_3$, and denote by $\langle T\rangle:= \langle u|T|u\rangle$ for convenience for any operator $T$.
Using self-adjointness of $L_3$, we have that $L_3 u = \lambda u$ where $\lambda \in \mathbb{R}$. And furthermore
$$ \langle L_3T\rangle = \langle L_3^* T\rangle = \lambda \cdot \langle T\rangle = \langle TL_3\rangle $$
which we can also write as
$$ \langle [T,L_3] \rangle = 0 $$
for any operator $T$. This implies
$$ \frac{1}{\hbar}\langle L_1^2\rangle = \langle -iL_1L_2L_3 + iL_1L_3L_2 \rangle = \langle -iL_3L_1L_2 +i L_1L_3L_2\rangle = \frac{1}{\hbar}\langle L_2^2\rangle $$
The first and third equalities are via the defining relationship $[L_i,L_j] = i \hbar \epsilon_{ijk} L_k$. The middle equality is the general relationship derived above, applied to the first summand. (And is precisely the identity that you said you could show in the question.)
Remark: it is important to note that the expression $\langle u| [T,A] |u\rangle = 0$ holds whenever $A$ is self adjoint and $u$ is an eigenvector for $A$. This does not imply that $[T,A] = 0$. This is already clear in a finite dimensional vector space where we can represent operators by matrices: consider $A = \begin{pmatrix} 1 & 0 \\ 0 & 2\end{pmatrix}$ and $T = \begin{pmatrix} 1 & 1 \\ 0 & 0\end{pmatrix}$. The commutator $[T,A] = \begin{pmatrix} 0 & 1 \\ 0 & 0\end{pmatrix}$, which is zero on the diagonals (as required), but is not the zero operator.