As far as I can determine; their estimates of new daily infections were, at least in part, based on people telling them they had had a positive test.
This was fine at the start, as every test was a test for current infection.
However, that is no longer the case, as many more people are getting antibody tests to test for past infection, but the app doesn't seem to have been able to differentiate; people are reporting positives but the app can't tell whether they are positive antibody tests (so not relevant for current daily infection estimates) or not.
So they have reevaluated their estimates for daily infection based, presumably, on removing a certain percentage of positives assumed to be antibody positives not current infection ones.
The upshot of this is that their estimate of daily infection has dropped significantly over the past week, but how much is actually a drop in real infection we don't know as they didn't reissue last week's number based on the new methodology.
They haven't redone their overall current infection number either, which makes me suspicious that it is too high. It's about double what the ONS surveillance survey would suggest.