For anyone who also has this issue, the best option at present is to to create a watchlist with the affected points, then export data in .xlsx format for the affected period.
Bad rows can be marked for deletion in excel by entering 'delete' in the modify column. The .xlsx file can then be imported via the 'data import' module page and the marked rows are deleted from the database.
On a MangoES, this took around 1-2 seconds per ten rows. If you try to process too much data (like many hundreds of rows) then Mango might crash and you may have to restart it (check with Infinite Automation for more info).