We present a systematic study of identified–hadron production (protons, \(\pi ^\pm\), \(K^\pm\)) in \(p{+}\textrm{C}\) at \(p_\textrm{lab}=90~\textrm{GeV}/c\) over \(\theta =0\)–\(80~\textrm{mrad}\) using PYTHIA 8.315 with the Angantyr extension. To match the NA61/SHINE production-event definition, we apply a generator-level selection that suppresses events dominated by a single forward, beam-momentum proton at very small angle (elastic-/quasi-elastic–like “beam survivors”), while leaving hadron production in accepted inelastic events unchanged. We find that both the default Angantyr baseline and the non-suppressed fixed-radii geometry underestimate forward proton yields for \(p\gtrsim 40~\textrm{GeV}/c\), despite describing mid-angle production reasonably well. To address this, we introduce an impact–parameter–dependent suppression applied after the fixed–radii ND/SD/DD/EL classification, modifying only the relative fractions of ND/SD/DD/EL NN encounter types while leaving particle production within accepted encounters unchanged. Relative to a b–invariant (“flat”) alternative, the b–dependent scheme hardens the forward tail without introducing unphysical distortions outside the forward region. Including modest hadronic rescattering enhances soft yields at larger angles while preserving the very-forward high-p component; with a common strangeness retune, the combined setup gives the most consistent description across protons, pions, and kaons. These results show that a minimal impact-parameter–dependent change in the mixture of NN encounter types can improve forward yields at SPS energies without distorting large-angle production, once the NA61/SHINE production-event class is reproduced.
