In the example given in tutorials/roostats/rs301_splot.C, the isolation variable is used to illustrate the effect of sWeight, but it's also included in the PDF used in the sWeight evaluation. In principle, the sWeight should be obtained from variables that are uncorrelated to the `control` variable, as stated in the explanation at the beginning of the script.
To visualize the issue, one can increase the statistics from 1000 events to 100000 in the example, and plot correct isolation distribution from PDF in the bottom two figures. This will result in the attached sPlot.png figure, where the discrepancy in left side of the bottom isolation variable is quite clear.
A modified version of this example to fix this issue is also attached in case it's useful.