sampled PAC transfer function to discrete step response

I am simulating an amplifier followed by 4 track and hold circuits, each sampling consecutively (interleaved). The duty cycle of each of the four clocks is about 20%

i am running sampled PAC, and i am choosing the sampling time of one of the four clocks, and i am looking at the output of each of the 4 TH, to see the effective BW of the amplifier+TH.

There were a couple of things that i cannot explain to myself

1) It seems that I have some ringing on the TF at the sampling frequency. Is this a memory affect of the TH? i see that the size of the ringing depends on the clock duty cycle

2) I expect the TH that i am using as the trigger clock to have a different response than the other 3 (clock injection makes them have different LF gain) but i expect the other 3 to have roughly the same response, but i see that the phase of the ringing is different between each of the other 3

3) Because of the ringing, i want to convert the TF to a step response with a inverse fft. since there is no built in function, i think i can take the conjugate, than put it through a regular fft, then take the conjugate again. the problem is that the fft is failing when i put in complex frequency data