You seem to have gone the long way round to collect frequency-domain data from your system.
It is difficult to suggest a specific approach, however finding some way to determine the amplitude and phase characteristics at each frequency would likely be the first step. Then use the procedures in Frequency-Domain Data in iddata to create your data object.
A simpler way would likely be to use the same sampling frequency, and record the respoinse of your system to an impulse, if you can do that. Then use that time-domain matrix, or the fft of it as a frequency-domain matrix to use with iddata (or idfrd).
