The big advantage of keeping one DTM is I guess that if at some point in the future I get better survey data (which is what I'd expect on my current project, so far only at outline stage with some of the site measurements being guestimated) then I simply replace the old "existing" data with the new but without losing the information in all the modifiers that convert it to "proposed".   In theory, anyway. I've not used the Site Model tool enough yet to have a feel for how reliable it is