I wonder how much of a sampling bias there is. It's plausible to me that users of non-IE11 browsers are more likely to be using ad/tracking blockers which block Google Analytics. That could skew the relative population of "users who ran Google Analytics" towards more IE11 users.

Were you able to use other sources than Google Analytics to look at the browser population? I imagine working with the NHS that things like HTTP access logs are harder to wrangle (if they exist at all)