Ha! Finally worked it out. In the include you have to include the full path for all of the directories - but NOT the url:
wget -r -A xls,xlsx,"" -l 2 -I /government/uploads/system/uploads/attachment_data/file/,/government/publications/,/government/collections/departmental-spend-over-500 https://www.gov.uk/government/collections/departmental-spend-over-500
Not obvious - well not to me anyway...
(scroll right in the code box to see it all)
Edit: Actually better - I've split it out here:
wget -r -A xls,xlsx,"" -l 2 -I /government/uploads/system/uploads/attachment_data/file/, /government/publications/, /government/collections/departmental-spend-over-500 https://www.gov.uk/government/collections/departmental-spend-over-500
Line 1: recursive, include xls & xlsx & files that don't have extensions (- in this case html files...) and do two levels from where specified in line 5
Lines 2-4: include these paths/directories from the top url (ie exclude everything else)
Line 5: where to start from