Ok so now I have developed calibration regression models for predicting the levels of 2 Fatty acid isomers 1 and 2. The isomer 1 is present in high concentrations than the isomer 2. I used 73 samples to develop the calibration method. The model with the higher isomer concentration had the higher RMSE value while the one with a lower isomer concentration had a lower RMSE value.

Based on the formula of the RMSE value, it can be seen that it is dependant on the residual directly and inversely related to the isomer concentration. So why is that (even in a published data we have the same trend) model with hiher calibration values has higher RMSE and those with lower calibration sample values had lower RMSE?