But if we only compare countries with similar GDP (per capita, presumably) to the US then the sample size becomes tiny and the US is an outlier on the religiosi

But if we only compare countries with similar GDP (per capita, presumably) to the US then the sample size becomes tiny and the US is an outlier on the religiosity scale. You seem to be trying to imply a correlation exsits between religiosity and violence based on a single outlier from a tiny sample.