Very important point! A battle I've had myself!
In ecology, field data often has small sample sizes, due to the work required
If we find a signal, eg in a regression of x on y, everyone is happy
If we find none, reviewers etc blame the sample size rather than accept that there may be no relationship between x & y in reality