Revisiting Model Inversion Evaluation: From Misleading Standards to Reliable Privacy Assessment