The predictions made by AlphaFold are now indistinguishable from experimental data collection error so folks aren't super concerned. Anyway structures are typically qualititaive tools useful for thinking about proteins, rather than direct targets of computational predictions (hasn't stopped people from trying).