“…For (4), is obviously pseudo‐effective because otherwise, will be ruled by horizontal (with respect to ) rational curves by (2), which contradicts that is non‐uniruled. The exceptional locus of a flip contraction is of dimension 1, so it does not intersect , neither does that of an extremal divisorial contraction because it is uniruled (see the proof of [, Lemma 3.2]). Running an LMMP for , by induction we get a needed map .…”