What the analysis says
What produces the ankle, instep, and suppression of the UHE cosmic-ray spectrum, over four members: H-9 GZK extragalactic protons, H-12 He/CNO source-composition, H-21 Galactic→extragalactic transition, residual H-37. Prior [0.4, 0.34, 0.16, 0.1] → posterior [0.6951, 0.2052, 0.0831, 0.0166]. Four per-instrument joints did the work: CG-2 (TA), CG-4 (Fly’s Eye), CG-5 (HiRes) favour H-9 because the suppression’s E_1/2 lands on the parameter-insensitive GZK-proton value; CG-3 (Auger) anchors on H-12 via the >5.5σ instep plus sky-uniformity. Three GZK-leaning witnesses outweigh one composition-leaning one; the residual collapses to ~1.7%.
What the model may not capture
The four CGs are scored as independent witnesses, but each instrument’s absolute energy scale is calibrated through the same air-shower hadronic-interaction models (QGSJET/EPOS/Sibyll) — a cross-instrument common mode the per-instrument partition cannot see. The trust caps (0.9, 0.9, 0.85, 0.7) fold in only each instrument’s own systematic, so a correlated model bias would shift the “GZK coincidence” the same way in three groups while the arithmetic counts them separately. H-9’s load-bearing discriminator — E_1/2 on the GZK energy — recurs in CG-2/4/5 as essentially one claim measured thrice, so treated as near-independent it over-drives H-9. What would separate H-9 (proton) from H-12 (heavy) is Xmax composition data, excluded at orientation (exclusion 3) — the true discriminator sits outside the model. And exhaustive_by_construction: false: the members share a single-dominant-account framing, but reality may be a mixture (propagation shaping a partly-heavy composition), which exclusive blocks cannot express and which is no less consequential than the listed answers.
What would help
Auger/TA Xmax mass-composition results — exists, unread (scoped out); moves H-9↔H-12 most. The hadronic-model energy-scale systematic shared across all four instruments — exists, unread. The Auger–TA composition reconciliation working-group output — exists, unclear whether it resolves the split.
Confusions and contradictions
AGASA once reported no suppression (S-49), later overturned — resolved, not live. The live conflict the analysis passes over: the spectrum agrees across experiments (what drives the cluster) while composition does not — Auger infers a heavier high-energy composition, TA is compatible with protons. H-9 needs protons, H-12 needs heavy nuclei. The 0.70 H-9 rests on spectral agreement that composition data actively contest, and the model reads only the spectrum. Shipped unfixed.
External consensus
No accepted winner; the Auger-weighted modern view leans mixed/heavier composition with source-imposed structure — nearer H-12 than the posterior’s strong H-9. The 0.70 sits above mainstream, largely because the H-12-favouring composition evidence was excluded by scope.