I did an error during compilation of Hadr07, which I now fixed. The output of v11.3 is only 292 F-18 with your file of PhysicsList.cc (attached some extract from the output), while from expeperimental data (attached) for 16 MeV protons we should get about 1392 F-18 in 2 cm long [O18]H2O.
I talked to an expert in this area (Pedro Arce) and he told me that there is aproblem with the use of evaluated proton data, which would be called in AllHP type physics list (I think for example QGSP_BIC_AllHP). The problem is long known (see DOI: 10.1016/j.apradiso.2017.12.009 Table 2, but that publication used GEANT4 v10.1, and the energy used is 13 MeV).
I hope that the GEANT4 can use the evaluated data more efficiently.
Regards
Can you tell in which GENAT4 version + patch number it was fixed? I may ask my friend to check output of Hadr08 with this simulation in version 11.2 (not sure which patch, but probably one of the last ones).
If you compare in output ProtonInelastic: in AllHP version with v11.4 reference there is about 8370 interactions, in candidate output protonInelastic= 8565, in reference there is protonInelastic= 4237.
It looks like in AllHP there is correct interaction of protons but cross section for F-18 production is zero. In candidate version the cross section was correctly (or almost correctly) used. The difference between candidate F-18 production (896 F-18) and the measured one (1392 F-18 per 1e6 protons) may result from difference between evaluated data EXFOR and TENDL (see http://doi.org/10.1016/j.apradiso.2017.12.009, Figure 3 b).