FDTD region: We recommend to leave at least half the max wavelength distance between edge of critical object (sphere in this case) and PML boundaries and then perform convergence testing. This means that as you are increasing the FDTD span, results should converge.
TFSF source: It should be large enough to enclose the particle. Also, since source requires ~2 mesh cells to inject the light (depicted graphically by light white shading that surrounds the source), you will need to choose it larger so that you can fit total analysis group (consists of 6 DFT monitors) inside it. The other thing to bare in mind is how you interpret the results of absorption and scattering. For more information on how to normalize the power with TFSF source, please take a look at this page: https://kb.lumerical.com/en/ref_sim_obj_tfsf_power_normalization.html
Mesh: Mesh override is recommended to be large enough to encompass TFSF. Please see the note on the link you provided on your post:
Note: Mesh override region
For simulations with metals, the mesh override region is often used to more accurately
resolve the locations of the metal interface, especially with curved surfaces. In this
simulation, the mesh override region is large enough to encompass not only the gold
sphere, but also the entire TFSF region. This was done intentionally; the TFSF sources
work best in uniformly meshed regions.
It is a part of convergence test to use finer mesh for the override mesh (and also FDTD mesh accuracy) to make sure the results do not change with the mesh size, specifically in this case where you have a sphere that has curved surfaces.
monitors: There are two regions here, total field area and scattered field area. Total filed that captures the total transmission, needs to be located inside TFSF source (should not be located in the white region of TFSF source as is explained above) and encompass sphere. Set of monitors to capture scatter field (scat analysis group) need to be located outside TFSF source.
We have also introduced two new commands, mie3d and mie3ds12 that you can use them to compare your simulation results with theoretical results.
I hope I could answer your question.