Transactions are posted in batches — one hash per bank import cycle rather than one hash per donation. A single hash covers as many as 54 ledger rows totaling $18,854. This is not a data integrity problem, but reduces per-transaction auditability.

The current automated scan flags 73 ledger rows across 35 potential duplicate-signature groups. One previously flagged entry, Seacoast Republican Women $40, now appears only once and is resolved. Three older exact duplicate groups remain. Newer review candidates include repeated Perplexity.AI $50 rows and July pairs for Residence Inn San Jose $313.13 and Uber $28.98. Some same-day/same-amount pairs may be legitimate separate charges, so they should be reviewed before any reversal is posted.

Two $2,500 payments to Late Apex Marketing are categorized as Administrative Expense. Based on the vendor name, these are almost certainly marketing expenditures and should be reclassified as Marketing Expense for accurate FPPC reporting. Reclassification would move $5,000 from administrative to marketing expense without changing total spending.

The latest batch introduces apparent spelling and capitalization variants that split the same donor or vendor across separate labels. Examples include Southwest AIrlines (12 rows) and SOuthwest alongside Southwest Airlines; Brita LIndstrom; Hyatt Regencey; and several Vid Muse variants, including "Xredits." These variants do not change total dollars, but they reduce the accuracy of donor/vendor grouping, search, and category drill-downs.

The dashboard follows the ledger API's published aggregate totals, as previously selected. Independently summing the 1,217 returned rows produces differences of $140.01 incoming, $222.51 outgoing, and $82.50 net. This does not prove that either result is wrong, but it means transaction count alone is not sufficient to verify that the dashboard and ledger are synchronized.