from the horse's (or rather report authors') mouth
There are, however, important limitations when considering the generalisability of our findings. Educational settings opened after national lockdown when SARS-CoV-2 incidence was low and only in regions with low community transmission. Settings that opened had stringent social distancing and infection control measures in please and, in addition to school attendance not being mandatory, there were strict protocols for class and bubble sizes, which may not be achievable when schools opening fully in the next academic year (and indeed, updated schools guidance now recognises that bubble size may need to be increased from September to ensure that a full range of activities is feasible). Only 1.6 million of the 8.9 million students nationally attended any educational setting during the summer mini-term. Additionally, very few secondary schools opened (and those that did, did so with small class sizes) during the summer mini-term and our results, therefore, are not likely to be generalisable to secondary schools, especially since the risk of infection, disease and transmission is likely to be higher in older than younger children. Moreover, each situation was risk-assessed on a case-by-case basis and only a few settings were selected for wider testing
And to be clear the 1.64m is for all educational settings and relates to the final day of the analysis period. On June 30th according to the data table that accompanies the report the number of DCs attending primary or secondary was 1.03m, and the average day over the analysis period for primary and secondary schools was 0.67m.
I write all the above with a DC due to start Y13 in Sep. I absolutely want schools back full time, but there are risks and to pretend this is not the case is symptomatic of the failing of the education system in this country for at least the last five decades, if not longer denying reality. I'd agree with Chris Whitty that the risks associated with not going back are though far higher.