Power conversion efficiency is at best the 3rd or 4th important factor here.

The first factor would be the hodgepodge of wall dongles one needs to own and maintain (plus the cost of buying a dongle for each device that doesn't have one, or multiple of them per device in case you want to charge your phone/laptop/etc in more than one location at home).

The second factor is the "smoothness" of your DC sources. Most of the common LED lamps have a pretty ugly signal shape, and not at all close to a DC flat line. This is mostly unavoidable as AC->Smooth DC conversion is more expensive than AC-> DC + a ton of 120Hz, 240Hz,... on top of it. So, common LED lights tend to opt for cheaper "electronics". People notice the flickery LED lights to various degrees (some get headaches, some outright see the flickers, some claim to be totally oblivious to the difference). The DC "quality" also affects some fairly sensitive electronic devices, so some AC->DC adaptors are fairly sophisticated. A central high quality AC->DC convertor (combined with DC wiring) has better scalability when you need to care about smoothness (it can be a basic quality of life matter for some people).

The third and fourth factors are power discipation and conversion efficiency. They are the same thing, with two remedies: more $ to remedy the inefficiency (which is really small these days, if you go for switching convertors), and plans for heat to discipate properly (devices end up with pretty hot adaptors).

It varies pretty wildly, often efficiency ends up being dependent on the load since power supplies usually get optimized for a certain load range. Individually, the numbers might not look too bad, but when you think about how many individual AC->DC supplies you have, the losses can add up.

I've been involved in a side project developing a consumer-friendly rating of "power quality" for AC devices and AC->DC power supplies which summarizes efficiency over a range of loads, as well as incorporating power factor measurements. We've been testing common devices such as USB power supplies for phones and such, as well as things like laptop power supplies, due to how numerous they are. We've had a few surprises, for example, Apple power supplies generally don't fare that well.

Power Quality Score: https://pqs.app/ Detailed test data is public for some devices but not all, since we're trying to find paths to revenue starting with subscriptions for full test results. Let us know if you have feedback.

You might also be interested in the Youtube channel of my friend/PQS collaborator, where he's done some "deep dive" videos of testing some of the devices in the PQS database- particularly AC->DC USB power supplies due to how ubiquitous they are now- https://www.youtube.com/c/AllThingsOnePlace/ .

I"m not sure I'm really the person to answer this but I would guess inverter and/or buck/boost converters are in the 90%-95% range. So, chaining DC -> AC -> DC gives about anywhere from 15% to 30% losses.

I think the better argument is one for reduced 'hardware complexity'. Instead of having an inverter that then goes through a rectifier, all you need is a buck converter.