Health related datasets are notoriously hard to work with - been there done that and it was bad enough with just one local authporities data. I would expect it to be far worse when you're collating it from multiple sources nationwide. Wasn't it yesterday or the day before that the Scottish data was initally left out and added in later due to processing delays?
Tbh, ignore the headline data on deaths in any case. Always look at the 'by date of death' graph on the dashboard as it gives a better picture of what's happening.