We plot here the most prominent parts of the precision matrices in Fig. 6, both from the mocks (blue, dotted) and from our non-Gaussian model (red, fit using the |$\mathcal {L}_{2}$| likelihood). The correlation function is evaluated in bins with δr = 4 h⁻¹ Mpc and δμ = 0.1. In these plots bins are ordered first in r (covering r = 102 h⁻¹ Mpc to r = 114 h⁻¹ Mpc on), then in μ (μ = 0 to μ = 1).