Why an earnings beat can be misleading when analyst estimates use different definitions, and how to check comparability before calculating consensus.