In the midst of doing some modeling work I discovered that the plots for the RTS-style tests were incorrect - the training AUC was used for the vertical dashed line instead of the metric that was actually used for the replicates. The p-values reported by the test were still correct, as were the values in the reps table; it was only the plots that were affected.
This issue has now been fixed on GitHub, and the fix will be part of the next upload to CRAN.