Can the assumption of a non‐random search improve our prediction of butterfly fluxes between resource patches?