Sample data

I'm using the data from the sample construction dataset and attempting to link invoices in CURRENT_ART_RECORD_1 (TINV) with invoices in ACTIVITY_ARA_RECORD_2 (AINV), but notice that there are two invoices under the USA customer where the ID digits mysteriously transpose between 01 and 96 e.g. 010030000002 becomes 960030000002. Any idea why this happens - is it bad sample data?