Yeah you don't want a linear reg. You need an actual buck reg- there are ICs to do this. It's somewhat difficult to design them for stability. Look at inductors carefully for efficiency, and be concerned with capacitor ESR.
You might wanna check and see how much the original transformers are drawing as no-load power from 110v. It may not be all that much. You need to measure actual difference in DC draw off the batts ideally, because a meter reading current at 110v directly doesn't take into account the phase relationship so power can be different than current * voltage.
The problem is the devices themselves may consume more than a few watts when "not doing anything" while still powered up. Eliminating the transformer will not help that at all.
OK, wait, I see you're concerned about the 110v inverter's quiescent current draw. Yes, some inverters take many watts just to stay on. I've seen cheap ones that take like an amp at 12v. Some retarded one don't put the fan on a thermostat and make the fan run all the time, for like a 50mA @12v load.
But, it varies widely. Unfortunately, mfgs often don't properly document the quiescent current, and disreputable ones don't document it ACCURATELY. I do have a 12v 2KW inverter which draws 80mA @12v which is awesome. This is really the way to go, honestly.
Be aware that a buck converter MAY draw significant quiescent current itself. Linear ones can easily be made to draw next to nothing- but, only 50% efficient. And your devices probably aren't 12v! Say you have a router that uses 200mA @ 9V. Linear regs will burn up 3.6W right there for that one device. Now if you used highly efficient buck reg to 12v, and did a simple linear reg to 9v, you'd only lose 0.6W- but it adds up.
An inverter which uses 100mA (2.7W) to stay on as quiescent current, but avoids the linear reg losses in the individual devices- may be more efficient in the end.