Notes. Results of the knife jacking method applied on the entire sample when removing and adding back objects in the given redshift ranges from the first column to build the composite template. The dust parameters are stable. Even though it is still consistent with the other ones, the results without one redshift range (namely 5.0 < z < 6.0) differ. This could be explained because the 5.0 < z < 6.0 range contains the widest wavelength range and it is therefore more constraining to estimate β_RJ and T_dust. So, we cannot rule out a small variation of these two parameters with the redshift inside the quoted uncertainties.