Great work. Very in depth as usual.
From my point of view and experience with layering, yes, 1 layer will def be better than no layers. Think of no layers being the same as 3 dimensional cubes representing the outline of the geometry. You would get results that are likely of significant error and unrepresentative of actual results. Having used 1 layer myself to perform quick runs with minimal headaches on very complex geometry, I agree with your conclusion and results.
As for the different between 1-4 layers. This is much more subjective. I do think some flow phenomena like down stream vortexes or any simulation involving turbulence effects down stream will be significantly influenced by the number of layers and expansion ratio. I believe that the small numerical changes associated with the varying number of layers only truly have an effect the further the influenced area is behind said flow changes that are influenced by the layers itself. Of course, there is a practical limit to it that will need to be judged depending on use case. On top of this, for a laminar or steady-state simulation, these effects would not apply.
Again, fantastic work on the layer generation. We can start recommending using a single layer which is much easier to generate than multiple layers for FSAE projects or other similar simulations.