In some new iteration of the ols_regress() command, it appears as in the "Model Summary" table, the Mean Square Error (and thus RMSE) is being computed as SSE/n instead of SSE/(n-k-1). See image below:

(An check is that the ANOVA table results for MSE should be identical to the table provided. )
It should be an easy fix, but there may be other computations that are misaligned here.