As far as I know, the cell size and cell height are directly responsible for the accuracy of the navmesh. A smaller cell size attribute leads to a much better defined navmesh am I correct? Wouldn’t increasing the cell size lead to a less accurate navmesh?
It would be great if we could feed it an already generated navmesh and use the Path finding functionalities later on. This would make it possible to use big meshes and have a detailed navmesh without losing anything performance wise. I think it would be worth it to take a look into it.