New Advances in Inference by Recursive Conditioning