In "M-theory on manifolds of $G_2$ holonomy: the first twenty years" by Duff, it is claimed (e.g. in section 8) that for compactification on singular 7-folds to be possible, we need to consider not the 11D supergravity (SUGRA) approximation to M-theory but "full M-theory". Such singular compactifications are desirable due to the absence of chiral matter in smooth 7-fold compactifications.
In contrast, many publications on M-theory compactified on 7-folds seem to just do Kaluza-Klein reduction of 11D SUGRA on the singular 7-folds, not considering "full M-theory" (as far as I am concerned, the M2- and M5-branes are part of 11D SUGRA as solitonic objects, maybe I'm wrong/non-standard with that view?). One example of this is "On gauge enhancemenet and singular limits in $G_2$ compactifications of M-theory" by Halverson and Morrison, where no "full" M-theory is in sight as far as I can see. There are many other such papers where the SUGRA approximation is the essential starting point for the Kaluza-Klein reductions.
So what, exactly, is meant by Duff's remark that singular compactifications are only possible for "full M-theory"? In what way does this compactification of "full M-theory" differ from a standard Kaluza-Klein reduction, and how does it allow for singular compactifications while 11D SUGRA only allows for smooth compactifications?