Different chi2 evaluate sampler cobaya CAMB vs CLASS

[Helena Garcia]

I am running the evaluate sampler in cobaya fixing all the LCDM and nuisance parameters of the likelihoods to the same values for both CLASS and CAMB (all of these parameters are set to the same values in both codes: logA,ns,ombh2,omch2,tau,A_planck,calib_100T,calib_217T,A_cib_217,xi_sz_cib,A_sz,ksz_norm, gal545_A_100,gal545_A_143 , gal545_A_143_217, gal545_A_217, ps_A_100_100, ps_A_143_143,ps_A_143_217, ps_A_217_217, galf_TE_A_100, galf_TE_A_100_143, galf_TE_A_100_217, galf_TE_A_143 , galf_TE_A_143_217, galf_TE_A_217, As,omegam,omegamh2,omega_de,Yhe).
I am also setting in CAMB nnu to 3.044 and mnu to and equivalently in CLASS N_ur to 3.044 and N_ncdm to 0
Adding Planck and BAODR16 likelihoods I am getting a difference in the chi2 of 3.36880000000019. The derived parameters zre, sigma 8, age, and rdrag are also different. I wanted to check that I obtained the same chi2 when running both codes with the same likelihoods but I didn't. Where can be the problem?
Thanks in advance!

[Antony Lewis]

Different thetastar/approx thetastar?

[Helena Garcia]

I am fixing H0 instead of theta, since the definition of theta for CLASS/CAMB is different.
The output .txt file doesn't provide theta_star, can I add something to the input .yaml file so I get the value od theta_star in the output and compare their value?

[Antony Lewis]

H0 should be fine, but it's not in your list of parameters.

[Helena Garcia]

Sorry, I missed to include it. H0 is also fixed to the same value in both .yaml input files for CLASS and CAMB. Any idea where is the difference in the chi2 coming from? Thanks!