It would be nice to separate out the AGN and SF components in the marginal plots

[eclake]

This would require performing the LSF convolution on the currently saved full SED outputs OR storing the convolved marginal for each separately at the BEAGLE stage.