That will be quite specific to the propellant. I've burned EX motors that left only scraps of the phenolic liner behind (very fast burn, very high pressure, very hot, practically invisible flame in daylight and barely detectable exhaust), and recently one slow burn that produces essentially zero slag due to composition. On the other hand, try burning something high in carbonates and metals (say, calcium carb plus strontium carb plus Mg for a nice color flame) in a small motor... lots of slag will be left behind. But burn it in a much larger motor, and the percentage mass slag left behind will reduce. You might also find the result different for vertical vs horizontal test stand orientation.
I'd argue that essentially whatever mass is lost during the burn is the "propellant", even if it started life as a casting tube, liner, nozzle, or smoke, and the resulting effective characteristic velocity is likely lowered due to the additional rather non-optimal propellant mass. This is partially traded off in slightly higher mass fraction.
It all depends on goals. For effects motors, performance mostly doesn't matter. For high performance, I like a metric something like density impulse divided by burn rate at target chamber pressure. That essentially tells me how much total impulse I can stuff within a given diameter with roughly the same safety margin. For our rockets of modest size, drag dominates over mass. Diameter dominates drag.
Gerald