Feature request: ensemble tree templates

Dear KNIMErs,

In principle I love the ensemble tree learner nodes, with one exception: you need to know a bit too well what you're doing! It'd be so much better to know which combination of settings mimics the behaviour of Breiman's randomForest implementation in R (but avoids all the I/O overhead and R's architectural shortcomings), or which modifications make it behave more like other flavours of the algorithm.

The node description suggests that it's very similar to the R implementatio, with only a few missing features. I recall, however, that once I tried the tree ensemble regression nodes in default settings, and the results was a horribly overfit model. My R one-liner, on the other hand, exhibited a very reasonable out-of-the-box behaviour. 

Has this improved during the last few releases? If it has, it would still be useful to have an implementation default which is as true to the R reference as possible - is presume the "currently under development" arning is still in effect?