[RM-6511] tplCombinationsolve: different results (evidence and pdf) for qsoSolver and galaxySolver

Created on 2021-05-27 16:40:28 by Mira Sarkis. % Done: 0

Running QSO spectra using tplcombinationsolve simultaneously for both (refer to the attached param.json for more details):

• galaxySolver,
• qsoSolver,
• using identical parameters for both solvers (no ism, no igm),
• using the same templates, i.e., one meanQSO + 4 components,
  leads to differences in the pdf and the logEvidences.

while we can expect 50%-50% G/Q classification, we get 0%-100% or 11%-88% classification for a subset of spectra

• I attach the list of the concerned SDSS QSO spectra
• and the resulting .hdf5 file

Output directory: /net/CESAM/msarkis/SDSS_qso/output_18May_6451_qso_meanqso_4Comp