I think the easiest and fastest way is still to just use the largest separation, like you alluded to, if the clump sizes are not vastly different among them.
Unfortunately, I don't think Chrono or DEM-Engine has a readily available tool for what you are after exactly. But a more intricate sampling scheme is out of the core functionality of DEM-Engine as well. I can imagine an algorithm that does that: You first randomly (or follow whatever policy of your choice) pick clump types and store them in an array, then you know the axes-aligned bounding box of all the particles you want to create. Then if you also know the cross-section shape of the region you'd like to fill, this problem is a classic 3D strip packing problem. Then you can solve it yourself or use some tools for that, e.g.,
py3dbp,
packingsolver. Again, this seems like a standalone problem to me, and we are probably not going to integrate something like that into the solver itself any time soon; so you have to integrate it yourself.