I wouldn't even trust the API with the numbers. The CSV and OBX published by a number of banks I used were always broken in some interesting but obvious ways. Minutes replacing months (someone mixed up %m/%M), values with cents appended twice for international transactions, unique IDs in OBX rolling over every month, and many others. Unless the same API is used for their legal reporting, I don't have a reason to expect it's better. Scraping the pages was way more reliable.