On open questions in the geometric approach to structural learning Bayesian nets