load intersect is now much quicker. however there might be unexpected behaviour in detect_gene_copies (or maybe elsewhere) if a segment is present several times in a feature (duplication within the feature). this case was not encountered in rice and human data, but it is technically possible.