For Falcon 9 though, the first stage contributes so much of the dV that it is not particularly feasible to land it at the launch site - by the time it separates, it's moving at several kilometers per second away from the launch site towards the horizon. That opens up the question of where the launch/landing site pair is bound to be. A ship-based landing system, such as a semisubmersible oil drilling rig (only ~$1B used!) with a landing pad built onto it, would seem to be close to optimal for the F9 specifically.
The Falcon Heavy's 2 side boosters, on the other hand, will separate much earlier in the flight - they are likely to be much more practical to land at the launch site, particularly for direct geosynchronous payloads that allow a lot more vertical / near-vertical burn time in relation to horizontal.
It costs about 8km/s to get into low orbit from an atmosphere-less Earth. From our planet it costs about 9.4-1.0km/s because of aerodynamic and gravity losses. Those aerodynamic and gravity costs, and the Hohmann Transfer from groundlevel to orbit (trivial for LEO, more for GTO) are the only parts of the equation one can address while burning vertically to stay over the launch site.
Assuming 10, If the first stage only has to contribute 2km/s, it can address aerodynamic and gravity losses (while thrusting vertically), and then leave the second stage to boost for the horizon and achieve full orbit all on its own.
At 4-5km/s first stage contribution, it doesn't look like this is possible unless the payload is very undersized, leaving the second stage with enough dV to go from 'rising out of the upper atmosphere vertically' to 'circular orbit' all on its own.
The mass is a free variable, a greater vertical component is practical in GTO (and I don't have the modelling skills to say how much), and the second stage may launch fully fueled for F9R contra to the existing pattern, so I can't be 100% sure whether F9R will be a practical return-to-launchsite option for lightweight GTO payloads.
The Falcon Heavy's 2 side boosters, on the other hand, will separate much earlier in the flight - they are likely to be much more practical to land at the launch site, particularly for direct geosynchronous payloads that allow a lot more vertical / near-vertical burn time in relation to horizontal.