Advances in random forest tuning and improvements in false discovery rate controlling procedures via test-specific covariate adjustments