But then you need the sample return component to have a propellant package that can wait on the Martian surface till the rover arrives. Solids and Hypergolics are probably fine but liquids are problematical, unless you're doing an in-situ fuel manufacture like Zubrin advocates, but this has its own weight and complexity issues.

Another detail is that if either half of a split mission lands inside a patch of terrain that cannot be exited/entered, or the wrong side of a terrain barrier, then the whole mission is compromised. To avoid this, we tend to get focussed onto "safe" (i.e boring) locations with limited interest in the gelogy and topography.

My forecast is, however, for a medium-to-long distance rover to be directly landed via aerobraking and rockets in mission year 1, and then a return lander to be again directly injected, but with significant effort for terminal guidance, in mission year 3.

My best alternative option is to land the whole damn package in one go, and have a short-to-medium range rover pop out locally and get some rocks and dirt, then load them back to the return capsule on the landed base all within a 30-90 day mission, and return in the same martian year while the transfer orbit is still near-optimum. It needs a heavier mission, but cuts out a lot of the "what ifs".

Oh, and why not have a small drill on the lander to collect a local sample AND emplace a seismometer? (yeah, I know it add weights and complexity, but I hate landing a dumb base station when we could be putting a string of long-term sensor stations like the Apollo ALSEP)