Example: Strictness of the first distinction

Two successive SWAPs have a diagram with cross-register dependencies but total channel $\mathrm{id}$. The first input has no influence on the second output. Thus scheduled ancestry can strictly overestimate actual influence.