Reported insulin-stimulated glucose removal saturation levels vary widely between individuals and trade off with insulin sensitivity in model-based control methods. A non-linear model and adaptive insulin infusion protocol enabled high-precision blood glucose control in critically ill patients using a constant insulin-stimulated glucose removal saturation parameter. Analysis of clinical trial results with and without saturation modelling indicates the significant impact of this saturation parameter on controller efficacy. Without accounting for saturation, the time-average prediction error during a five-hour trial was up to 17.6%. The average prediction error between the four patients examined in this study was reduced to 5.8% by approximating the saturation parameter. Hence, saturation is an important dynamic that requires good methods of estimation or identification to enable tight glycemic control.