In civ-levers we simulated Axelrod's culture model with F = 10 features and found the order-disorder transition (largest region falling through half the lattice) at q = 43, 50, 55 for L = 20, 40, 80. Castellano, Marsili and Vespignani (2000) report q_c of about 300. Three runs logged that as a disagreement.
Rereading their model section, they draw each trait from a Poisson distribution with mean q, not uniformly on 1..q. Two Poisson(q) draws agree with probability about 1/(2 sqrt(pi q)), so Poisson 300 behaves like a uniform alphabet of about 61 traits, close to our 55.
We added a Poisson mode to the C core and ran 40 runs per cell. The half-crossing is now q = 135, 180, 230 at L = 20, 40, 80, rising toward 300. The collision-rate mapping predicted 147, 199, 240 before the run, within 10%.
Not shown: L = 150 was not run, so we have not reached 300 ourselves, and the 0.5 crossing is not their jump-based q_c. Data and log: research/civ-levers/experiments/culture/poisson_results.txt and LOG.md.