Apparent violations of Onsager reciprocity continue to be reported in models of transport through charged membranes and concentrated electrolytes, with asymmetric cross-coefficients frequently interpreted as evidence of broken microscopic reversibility. We develop a thermodynamic framework that clarifies when such asymmetries represent genuine non-reciprocal behaviour and when they are artefacts of modelling and data reduction. Using a charged-membrane cell model as a case study, we formulate entropy production in terms of conjugate flux–force pairs, construct the full symmetric Onsager matrix, and then analyse how variable reduction, spatial averaging, and concentration-dependent parameterisation affect the resulting reduced kinetic coefficients. Numerical reconstructions of published cross-coefficients, together with controlled examples based on explicitly symmetric Onsager matrices, show that nonlinear, concentration-dependent reductions can generate pronounced and systematically ordered separations between effective coefficients, even when the underlying phenomenological matrix remains exactly symmetric. We further distinguish equilibrium linear-response coefficients from finite-amplitude, steady-state effective parameters and discuss the impact of temporal memory effects in concentrated electrolytes. Taken together, these results demonstrate that the reported asymmetries in charged-membrane models do not constitute a violation of Onsager reciprocity and provide practical criteria for testing reciprocity claims in complex coupled transport systems.