I don’t get the point, here. Do you mean that the standard deviation of the errors of the Fortran library is lower, but the distribution of the errors have longer tails ? I’m asking because from the histograms I would say the standard deviation of the Julia library is actually the lowest one… But the total number of counts doesn’t seem to be the same between the txo histograms, so there’s something unclear to me here (I’m not even sure what is exactly “eps” here)…
EDIT: OK I missed it was a log vertical scale. Would be good to have the normal scale as well, and the calculated standard deviations of both histograms. I also notice that the Julia distribution is biased toward positive errors (but again to which point is difficult to evaluate on a log scale). The extracted mean values would be useful (and the standard deviations calculated for both an assumed mean at zero and the calculated mean).