Those routing instructions derive from a 'basic scenario' outlined in the DM3200 manual. They assume you've got 16 dedicated outboard inputs (ie: mikes, preamps, etc), and all 16 are in use simultaneously. Think: big band, orchestra, live recording, etc.
For DIY-ers - those who layer, or over dub a single instrument or vocal at a time - that routing scenario isn't particularly useful and wastes an entire mixer layer. So, instead of routing ProTools channels (inputs to the DM) to 17-32, change that to two layers: 1-14 and 17 -32. Leave one stereo pair (I use DM inputs 15/16) for tracking. (You can choose any pair, but if you've got your FireWire card set to "16" instead of '32" to save PC power, 15/16 is a good a choice as any).
As to recording back to ProTools: you've got several choices. I would suggest your best option is to use PT for bouncing your entire mix to a PT stereo track. You can use DM channel's 15/16 for that purpose.
Now - this is personal preference, but except for certain cases, I don't like 'printing' efx (verb, compressor, eq) onto tracks because that's the 'destructive' approach. This means if you don't like what you did when it comes time to mix, you've got to go back and re-do that track. I think it's better - allowing greater flexibility - to simply use the DM's onboard capabilities by non-destructively assigning the input channels to the DM's Stereo Bus. That allows you continual opportunity to adjust, change, abandon, or combine the DM's capabilities with PT plugins - which are also inserted non-destructively as well.
Regarding sub groups: let's tackle that question after you get your board set up for your particular workflow.
Hopefully some of this makes sense.
CaptDan