At this moment we are treating the barycenter as a legitimate position-reconstruction algorithm, and we have the choice of running the recon with either the barycenter, the eta algorithm of a NN.
I am wondering whether it would make sense to always calculate the barycenter coordinates, and then add two other columns with our best estimate of the photon position, be it using the eta function or any other algorithm.