Second Order Modelling of Incompressible Turbulent Buoyant Flows: Some Comments on the Maximum Principle