The basic ideas of Sarproz were firstly conceived during my phd thesis in the years 2003-2005 (check my thesis on Amazon or the urban target recognition paper). At that time I was using ERS and I calibrated the dataset with some corner reflectors. Since then, the calibration constant has been left in the code, and it is used when you process amplitude time series with the physical model.
If you use TSX data, surely the calibration is not correct, so the RCS value is just for reference but not calibrated.
Therefore, you should use the Reflectivity Map instead of the RCS. Even if not calibrated, at least you have an idea of its range of values.
In any case, the Reflectivity value tells you how strong is a target along your dataset. But it’s not a good indicator of the target stability. Better using an Amplitude stability index or a coherence-related index.
In any case, which indicator and which value to use depends on the particular site you are analyzing (sensor, number of images, characteristics of the terrain).
Try and make some experience…