We discuss utilization of TMU (Total Measurement Uncertainty) analysis based on Mandel Regression for scatterometry model referencing. We demonstrate practical instances where the reference metrology uncertainty seems to exceed that of the scatterometry model which, in turn, forces the TMU analysis into an invalid regime. Knowing that the source of this result is the wrong estimation of the reference uncertainty, we focus on reducing the error of the reference metrology as well as improving the reference uncertainty estimation. We are looking at practical aspects of reference metrology hardware and recipes as well as the whole reference metrology setup. We discuss in detail CD-SEM and AFM metrology, as well as other means used to qualify OCD (Optical Critical Dimension) models in a production environment.