This shows a difference in the software setup. The R estimator returns values below 0 which cannot exist, the BKDE estimators are limited to the range of the data by a choice made when writing the problem specific code for Bayes4. Other than this the plots are all similar.