I think Ansoft HFSS is good for metamaterial simulation but its not FDTD based . Empire Xccel is also good for metamaterial simulation and it is also FDTD based .
I would suggest CST, but since you are looking for an FDTD solver I would also go with the Lumerical choice. It has been 4 years since I last used RSoft. I was experiencing lots of trouble with 3-D problems in RSoft back in those days. Have they come up with a solution?
I am an experimentalist and I use Lumerical to play with, because it is very user-friendly. Several of my collaborators are theorists and they use RSoft. There is also Comsol, but we never really used that one.
I have used RSoft for metamaterials but their code isn't the most efficient. If you can deal with a higher learning curve, the open source meep and mpbc from MIT both work great for metamaterials. meep is FDTD while mpbc is a mode solver specifically for finding the allowable modes of periodic structures.
What sort of device you are working with using Metamaterial & in which range of frequency? Depending on your application, you can choose between Lumerical and XFDTD. I personally feel that XFDTD is suitable for most of the problems in RF, Microwave & mm-wave domain for Radiation, Scattering, cloaking etc. Lumerical is suitable for THz & optical applications.
For calculating the S-parameters, you need to introduce some reference planes (ports) to distinguish between the scattered field and the total field at those planes. At the context of FDTD, you might have an exact knowledge of the incident field at those ports. In addition you might need to use numerical calibration techniques. I don't know about EMTL and openEMS, but HFSS offers a plenty of different ports and is usually accurate enough for meta-materials, though it is not based on FDTD...
CST studio suite uses FDTD (together with other simulation methods) and can be used to compute S parameters for metamaterials, using periodic boundary conditions.