Why RFECV visualizer from yellowbrick finds number of optimal features more than number of features in train set?

Question

I am trying to perform feature selection and to use RFECV from yellowbrick.model_selection. I have 48 features in my train set. But when I run the code in below, the visualizer finds number of features as 49.

xgboost_base_model = xgb.XGBClassifier(tree_method='hist')
visualizer_xgb = RFECV(xgboost_base_model, step=3, cv=3, scoring='roc_auc')
visualizer_xgb.fit(X_train, y_train)
visualizer_xgb.show()