The tensor product (or the "spin sum") of two irreducible representations of spin $j_1$ and $j_2$ decomposes as the direct sum of irreps of spin $j=|j_1-j_2|$ to $j=j_1+j_2$. In other words, when you sum two spins $j_1$ and $j_2$, the total spin can take values $j=|j_1-j_2|, |j_1 - j_2| + 1, ..., j_1+j_2$. Each of these irreps of total spin $j$ host a multiplet with $J_3$ eigenvalues $m=-j, ...,j$. You can think of each of these $j$ as representing different orthogonal subspaces, in which we diagonalize $J_3$ with eigenvalues $m$ going from $-j$ to $j$
For example, here you want to take the tensor product of two spin 1 irreps $j_1=1$ and $j_2 =1$, so $j=0,1,2$ runs from 1-1=0 to 1+1=2, and for each $j$, $m$ runs from $-j$ to $j$. This is depicted in your image. There are no dots placed on $m=2$ $j=1$ because $|m|\leq j$ in our construction
Let's take another example to confirm our understanding, with $j_1 = 2$, $j_2 = 1$, then $j=2-1,2,2+1$ which is to say $j=1,2,3$ and
For $j=1$, $m=-j,...,j$ which means $m=-1,0,1$
Similarily for $j=2$, $m=-2,-1,0,1,2$
For $j=3$, $m=-3,...,3$