I am sure I am not including the compilation time. I run the script 2 or 3 times and the time I quoted is pretty consistent over all the runs.
I actually know the package BenchmarkTools.jl and it is doing (almost) the same thing that I am doing in this simple script. It is easy to isolate problems in simple scripts.
Of course time taken by my Fortran code includes the same number of runs and same number of orbit time-periods.