If p = mfBound M, then the p-th syzygy of M, which is computed by highSyzygy(M), should (this is a conjecture) be a "high Syzygy" in the sense required for matrixFactorization. In examples, the estimate seems sharp (except when M is already a high syzygy).
The actual formula used is:
mfBound M = 1+max(2*reven, 1+2*rodd)