When you bring a mesh into modo to build new topology you’re doing more than just changing the mesh into Tris. You’re removing edge loops that aren’t really needed anymore (ones that don’t really do anything for the silhouette). You’re also re-routing edge flow and collapsing edges, etc.
Characters or static meshes usually benefit from a retopo.
Static props can vary on seam requirements. If the mesh is truly static, it can use an object spaced normal map, but it should never be moved…or effected by any type of physics, etc. If it is going to move then it should use a tangent spaced normal, but that really all depends on the target engine.
We’re talking different types of seams…well, kind of. The fewer UV seams you can have in a game engine, the better it will run. The fewer seams you have in your UV’s also effect how well your normal maps will work. The diffuse textures won’t have seams (or very few) but that doesn’t directly effect your normals. So we’re talking the same and different seams.
xNormal doesn’t project any better, or worse, I just find it easier. The maker of the software has also released the SDK for it, so our programmers in house can create a custom tangent base (from our engine) to make sure that xNormal is reading the tangents of the vertex normals the same way the engine does…that way the bake in xNormal will match exactly in the engine. It also bakes crazy fast.
Normals, cavity and diffuse are baked from the HP to the LP in xNormal. AO in modo.
The unwrapping tools in modo are great (but a free plug-in from seneca is very helpful). Some people like headus UV ($100), roadkill ($0), it’s really up to what you like, and what makes you the most productive. They really all do the same thing, so it doesn’t really matter.
Don’t worry about the questions, you’re asking the right questions so I don’t mind answering them.