Reverse micellar solutions of soybean lecithin in a number of organic solvents show a dramatic increase of the shear viscosity after the addition of small quantities of water. Furthermore, a sol-gel transition is observed when the concentration is increased above a critical value, Ф*, of the volume fraction. Such a phenomenon was traditionally interpreted in terms of the establishment of an entangled network of worm-like aggregates. The idea is supported by theoretical models that, in a mean field approximation, suggest for a micellar size distribution function shaped exponential and for a mean micellar length, <L>, scaling with concentration following a law <L>=Ф^1/2. In this paper a review is presented of a number of experimental results that well agrees with such a working hypothesis. However, at least the results from an Incoherent Quasi Elastic Neutron Scattering and a Small Angle Neutron Scattering experiments are not interpretable within the proposed polymer-like approach. The observed discrepancies between the theory and the experiment strongly suggest for a critical revision of the currently accepted ideas. It will be shown as some erroneous indications, given by the theory, are originated by a misguided attempt to extrapolate results from mean field approximation to semi-dilute or dense systems. In particular, it will be shown how the current idea of a monotonous growth process of <L> with no respect ofthe concentration is in principle wrong. It will be shown how the experimentally observed growth process of the micellar aggregates is not rationalizable as a linear process. On the contrary, the observed dependence of the mean micellar size by the concentration implies a change in the system polydispersity, driven by inter-micellar interactions. Such a behaviour is an emerging complex property of the system, originated from the competition among all the species (aggregates of different size) existing in solution and resulting, in the semi-dilute regime, in a lower configurational entropy than mean field approaches would indicate. From this different point of view, it is possible to formulate new models able to fit also the data strongly in disagreement with the prediction of the conventional theory.