Exploring the limits of Hierarchical World Models in Reinforcement Learning