In this paper we study the multicast routing problem in all-optical WDM networks with sparse splitting capacity. In the literature, there are several proposals on finding the best way to construct multicast light-trees (or light-forests) with the objective of minimizing the number of wavelengths (link stress), the maximum end-to-end delay from the source to the destinations (maximum delay) and/or the total number of wavelength channels used (total cost) of the light-forests. However, balancing all these criteria is very difficult. Particularly, end-to-end delay and link tress cannot be minimized simultaneously. Hence, it is interesting to find a good approach that can provide a trade-off solution. Putting this into practice, we propose a comparative study of the most known algorithms and introduce a new one which can provide a good trade-off among those three criteria. Simulation results and comparison point out that our proposal provides the best link stress, a low maximum delay and low total cost of the multicast forest among considered algorithms. Especially, our proposal works best in dense networks, and/or with a large multicast group size in comparison to the classical algorithms.