Hi Ezra!
Unfortunately I don’t think there is a way to implement exactly what you are hoping to achieve in either of the census interfaces. The only additional thing I might suggest would be to look into using the custom node label functionality. This would allow the two node types to look distinct on the Sociogram so you could easily guide your participants with a prompt that would ask them to connect only nodes of different types. Again, this is a different approach to edge generation than the traditional dyad census but sounds like it will be much more suitable for what you want to achieve.
Tagging my colleagues here in case others have ideas to mention: @Joshua @caden