as far as I know, there hasn't been work (within the collaboration) on specific spectral features since BPZ uses templates. I'd say that depending on the spectral type of the galaxies and their redshift, the most informative feature will change. Therefore, using templates avoid biases related with putting too much weight to specific features. With respect to previous multifilter surveys, like COMBO17, I would say that what made very powerful the J-PAS filter system is that the design of the filters has been done from scratch to maximize the photo-z determination (this is explained in Sect.2 of the document). The result is that the filters are narrow (145A), almost top hat with no wings, and there is no gaps between them, covering the whole optical range.