Directionality of coupling from bivariate time series: how to avoid false causalities and missed connections.