The radial electric field Er, and particularly its gradient, has been invoked by various theories and empirical models as a crucial parameter for determining the transition to high confinement regimes, such as the onset of an Internal Transport Barrier in the plasma core and of the H-mode pedestal at the plasma edge. Following a recent absolute calibration of the charge exchange diagnostic system, we have evaluated the uncertainty on the calculated Er. Starting from the neoclassical moment approach of Hirshman and Sigmar, simple approximations have been used to reduce the full matrix calculation to a set of analytical formulas, adapted for the 2D toroidal geometry of JET to describe all collisionality regimes (banana, banana-plateau, Pfirsch-Schlüter), and to include a calculation of the error bars on Er. Here we compare this analytical calculation with the results of the JETTO and NCLASS codes. Specifically, we assess how different approaches to treat numerically certain input plasma parameters for this calculation (ion density and effective charge profiles) can yield very different results for Er. This is particularly clear for the plasma edge, where the contribution of the toroidal rotation velocity to Er becomes small and comparable to the poloidal velocity and pressure components: hence uncertainties in the ion density profiles dominate the calculation of Er. On the other hand, excluding such edge and core regions of the plasma (where the typical scale lengths become comparable to the ion poloidal Larmor radius), we find a striking similarity in the shape of Er in L-mode and ITB plasmas, and we demonstrate the role of prompt fast ion losses when comparing H-mode plasmas with forward and reversed ion ∇B-drift direction. Our analysis points clearly to the need for routine measurements of the poloidal velocity and Er if detailed comparison with code predictions and empirical scaling laws are to be made for studies and modelling of improved confinement regimes.