2(a) It is not sufficient to cite Lemma 20 in the notes which says
that Ford-Fulkerson runs in O(nmU) time, because big O notation means
that Ford-Fulkerson runs in at most n*m*U time (up to a constant
factor). You need to actually exhibit a run of FF which does actually
take at least M iterations by adapting the argument for the example on
page 36.

(b) Remember we are running Edmonds-Karp in this part, which uses the
augmenting path which has the fewest number of edges at each
iteration. Many students used the path s->a->b->c->d->e->f->t on the
third iteration instead of the shorter path s->a->b->c->t (or
something similar, depending on which path was chosen on the second
iteration).

4(b) It is not enough to define g(e) = f*(e)-f(e) because this g
function could have negative value on some edges, which would not be a
flow.