I think you may run into two problems:
– you can’t route groups to other groups
– you will run out of mixes if you try and do it using groups and then into matrices
Is there a reason you need to subgroup the drums/instruments/vox before grouping them again to the band group? In our church application we route all instruments directly to the BAND group and then use DCAs to control levels of multiple inputs (so although all the routing is to the BAND group, we have granular control of drums, vocals etc via DCAs).
Unless you need to do some processing or routing specifically on those sub-sub-groups, DCAs are your friend 
