In the BBKS formula in CCL the definition of "q" seems to be missing a factor of (T_0/2.7K)^2 [Eq. 3.9 Sugiyama 1995], anyway, we create a CCL "compatibility mode" for our BBKS which uses the same definition of "q" as CCL and the tests went through, we agree in 5 x 10^-5 (which increases as we include more knots in CCL splines). In these tests we also included the comparison between basic background quantities as Omegas, H/H0 and growth function. (We can breakdown these number in the future to get the largest error in redshift per observable)
from @vitenti