As you have a known waveform and so know what the FFT should be you should easily be able to experiment with the parameters to see what works best. I don't use Eldo so don't have the same options as you are seeing.
In general, more points is better - it just takes longer to calculate.
Rectangular is good if the waveform you are looking at is exactly periodic over the simulation time so a 1kHz waveform over 10ms for example. I think it needs the reference frequency to be specified. In general you would use one of the windowed functions (Hanning for example).
If you know the fundamental frequency, include that - I think that is the "reference frequency" in your window. It will give better results.
In general you would not "use datapoints" for the FFT - it needs resampling. So, use Cubic or Uniform. Again, Cubic is better but takes longer.
I am not sure what your "periodic" does but as your waveform is periodic it would seem a better option than "symmetric".
Have a look at your "advanced" tab as well (and read the Eldo manual!).
As I say, with a known waveform you have the perfect method to experiment and see the effect of the parameters because you know what the results should be - i.e. not what you have.
Keith