Yeah, 'absolute mass' as a standalone metric isn't super useful, but obviously maximizing payload mass is <ahem> massively important to a commercial operator because their sole reason for existing is to make profit. Any of those operators would tell you an obvious method of measuring is $/kg per revenue generating payload.
That said, the current metric of choice in space circles is actually Data/Dollar...which of course is really just a function of $/kg of payload (implicit in Data/Dollar is all program capex/opex, not just the on orbit asset, if we're being concise). The previous metric was usually more along the lines of transponders/dollar or transponders/kg, but again--as I hope is clear--all these metrics are basically all variations on the same theme. The more revenue generating gizmos you cram into the top of a rocket, the more money you're going to make.
This is no more evident than the seemingly counterintuitive dual-propulsion design we see in bigger GEO satellites, where they have both a chemical propulsion subsystem and a completely separate [much more mass efficient] electric propulsion subsystem. The latter of course adds quite a bit of cost to the asset, but overall it decreases the propulsion mass (hardware + propellants), explicitly so mass can be shifted to the payload. As you say, someone is basically paying the same price for the rocket regardless how much mass it launches (that's not quite true but its close enough), and in a similar fashion to the exercise I described in my last post, adding the EP system decreases the total $/kg of the revenue generating payload. Win.
And yes, you're absolutely right in noting the need to earmark mass for on-orbit and decommissioning propulsion (actually commissioning too--sometimes commercial satellites do initial testing in a different orbit than their service slot in the GEO arc), which is exactly why those EP systems are added. They're kind of terrible at orbit raising, which prefers high thrust, but are a very mass efficient way to perform on-orbit stationkeeping maneuvers that do just fine with the very low thrust from an EP system.
After Data/Dollar, ROI (more typically IRR, but I digress) is usually more important to a satellite operator than total profit. One of the biggest reasons for this is the inherent risk with an on-orbit asset: The longer the thing is up there the higher the risk of component failure, which of course can lead to a reduction in service, which of course means a reduction in revenue. This is typically a pretty well understood curve; statistically you know how many cells and strings are going to crap out on a solar array by EOL, for instance. Another factor in ROI over total profit is technology advancement. As the years pass The Jones' are going to put up something bigger and better than your satellite, and especially as space transitions more and more to raw data, its going to be increasingly easier for your customers to jump ship to someone with better data rates.