N-15 CSA-N-15-H-1 dipolar cross-correlation (eta(xy)) in proteins has been treated thus far with the model-free (MF) approach, where the global (R-C) and local (R-L) motions are assumed to be decoupled as a consequence of R-L much greater than R-C. In the context of eta(xy) it is additionally assumed that the local motion is very fast and highly symmetrical. We have recently applied to auto-correlated N-15 spin relaxation the slowly relaxing local structure (SRLS) approach, which accounts rigorously for mode-coupling. SRLS can analyze eta(xy) for arbitrary time scale separations between R-L and R-C. Simulations of eta(xy), for slow local motions are presented herein for the first time. Experimental eta(xy), values of RNase and AKeco could not be reproduced from best-fit parameters generated by data fitting that used axial potentials. Calculations showed they are reproducible using rhombic coupling/ordering potentials. The conformational exchange term, R-ex, "absorbs" potential rhombicity when axial potentials are used to fit the data. N-15 CSA variability and R-C anisotropy are shown to have a small effect on the analysis. The shape of the rhombic potentials detected corresponds to nearly "planar YMXM ordering" (M denotes the local ordering frame). This potential form is consistent with the known geometry of the peptide plane, the SRLS dynamic model in the R-perpendicular to(L) approximate to R-C regime, and makes possible associating the local director with the C-i-1(alpha)-C-i(alpha) axis. It is shown that back-calculation of NMR variables from the best-fit parameters and comparison with the experimental counterparts is an effective tool for identifying inappropriate elements of the dynamic model and can thereby help improve it.