Why might a critical residue with a high SASA be more likely to be a false negative by protein rigidity analysis than a critical residue with a low SASA