Position embedding is added inside each block? #2

DiT is to add positional embedding once before the transformer block. And according to Figure 3 (c) of FiT's paper, it seems that the position embedding is added inside each block. Is this an unfair comparison?