We address the question to what extent JIMWLK evolution is capable of taking into account angular correlations in a high energy hadronic wave function. Our conclusion is that angular (and indeed other) correlations in the wave function cannot be reliably calculated without taking into account Pomeron loops in the evolution. As an example we study numerically the energy evolution of angular correlations between dipole scattering amplitudes in the framework of the large $N_c$ approximation to JIMWLK evolution (the "`projectile dipole model"'). Target correlations are introduced via averaging over (isotropic) ensemble of anisotropic initial conditions. We find that correlations disappear very quickly with rapidity even inside the saturation radius. This is in accordance with our physical picture of JIMWLK evolution. The actual correlations inside the saturation radius in the target QCD wave function, on the other hand should remain sizable at any rapidity.

12 pages, 11 figures