Because "large" here is still "very tiny". Large with respect to "p-values" isn't the same as "actually large". Equating the "positive results" to "winning the powerball" is not a fair comparison. If there were Powerball-level results here we wouldn't need further studies to confirm the effect exists.