I've been exporting SNP-level H4 for downstream analysis. It seems that's not the correct approach and instead I should be using the summary-level H4 instead to define colocalisation.
actions:
run_coloc.R should output a TSV with the summary-level H4 for each gene in each locus. This should be used for downstream summaries, not the SNP-level H4 which I was using.