I strongly suggest you don't use the geometry made in radiant and converted to .obj as anything else except reference for position and scale.
The cleanup is a nightmare and the amount of work needed to connect all the edges rarely makes it worth it.
Also when i would build something i used a "beacon" object. Usually it's terrain or a road. It can be very difficult to align to objects in radiant, grid snapping and a lack of snap to object feature make it very challenging even when you hide all the unwanted bits and pieces.
So lets say you got a map like Ryiadh, big terrain with some buildings, there are cuts in the terrain where buildings are placed. In max that building is easy to position cause you can snap to vertices and edges and what not.
In radiant it's hard to see what you're doing cause of too many lines. So in max you'd position the building, export terrain to radiant. Then from max export the terrain+building. So when you import it in radiant it's easy to realign it cause you can just use the corner of the terrain.
Then go back to max, delete terrain and export just the building, now when you refresh the models in radiant the building is perfectly aligned with the terrain.
It's a bit of work but the alignment is perfect and you avoid tiny gaps between objects.
Also when exporting a more complex object like a buidling, you can't vertex lit of lighmap it all in the same time. Water drains, door knobs, flower pots, and similar curvy details with smaller polygons lightmap horribly, so split the object into 2 parts.
building_V (bits you want to vertex lit) and building_L (bits you want lightmapped) and then export them as 2 models.
You might need 2 or more building_L models, depending if some parts require lower _lightmapscale to make the shadows look better.
Basically it's walls=lightmapscale 1-.5, smaller objects like a chimney with a rounded edge=lightmapscale .25