If I understand you right, the graph you are showing is the result of a power spectrum computation. If this is the case, you have to be aware that the frequency resolution (df) of your result will be the inverse of your signal length (in seconds). So if you, for example, measure 100 ms of signal, the resolution on your graph will be 1/100 ms = 10 Hz.
Since you only have excitation with 400 Hz spacing, it means that only 1 out of 40 values (400 Hz / 10 Hz) of your result will be valid. the 39 other values will be (near) zero.
So to display your response as a "continuous plot" you need to extract the valid values of result. Extract the array of your result cluster and then extract every n values (n depends on your own measurement conditions as explained above).
Then re-insert your array to your cluster AND remember to also change the "df" value (frequency resolution) to "n*df", since your new frequency resolution is now n times the old one.
Alternatively if possible try to make your measurement length equal 1/400 Hz = 2.50 ms.
