We give a partonic interpretation for the deeply virtual Compton scattering (DVCS) measurements of the H1 and ZEUS Collaborations in the small-x B region in terms of generalized parton distributions. Thereby we have a closer look at the skewness effect, parameterization of the t-dependence, revealing the chromomagnetic pomeron, and at a model-dependent access to the anomalous gravitomagnetic moment of nucleon. We also quantify the reparameterization of generalized parton distributions resulting from the inclusion of radiative corrections up to next-to-next-to-leading order. Beyond the leading order approximation, our findings are compatible with a 'holographic' principle that would arise from a (broken) SO(2,1) symmetry. Utilizing our leading-order findings, we also perform a first model-dependent "dispersion relation" fit of HERMES and JLAB DVCS measurements. From that we extract the generalized parton distribution H on its cross-over line and predict the beam charge-spin asymmetry, measurable at COMPASS. This is nothing but the RDDA model in the limit b → ∞; practically, a large value b ≫ 1 is sufficient. This ansatz implies that the skewness function is set r(η/x) = 1 for all x. With such an initial condition, evolution, starting at a rather low input scale, will rapidly lead to an increase of the r-ratio. Thus, this GPD model fails to describe data, too. We will not go into details here, 7 Usage of term "dual" was motivated by the fact that in dual models [98] the s-channel amplitude is described by the t-channel exchanges. We add that this feature is more general and arises from crossing and the Sommerfeld-Watson transform of the t-channel SO(3) partial wave expansion. In Regge theory/phenomenology the resummation of t-channel exchanges is encoded in the Regge trajectory.