The title mostly explains it;
- CBC-MAC variable-length weakness with MAC size greater than the encryption block size
I've seen that by having the pairs $(M,T_m)$ and $(N,T_n)$, we can generate a third message $M'$ whose CBC-MAC will also be $T_n$ by simply doing $M'=M \mathbin\| ((n_1 \oplus T_m) \mathbin\| n_2 \ldots n_x)$
My question is how does this work when the MAC is of greater size than the block size used during encryption? In my specific case, the MAC is 20 bytes while the encryption blocks are 16 bytes. So far I have not seen any ways of dealing with this