is impedance matching vital ?

dear all,

i have some troubles with impedance matching, i do not know really if i need to do it or no,

this is my special case:
i designed a gm-c filter, the input of my DUT, is transconductor so my input impedance of the DUT is very high, so at the input port i lose no voltage, i am not concerned by current since my filter is supposed to process voltage not power.
at the output i put a buffer , which is a transconductor as well, so i can load the 50 load without any degradation on my filter response

simulation results with s21 gave me the expected shape, (actually gave me 6 dB gain, actually this is due to the definition of Pa which surmise a matching at the input)
so in my case i wonder if impedance matching apply ? although my colleagues keep saying it's not correct, without giving any convincing argument

as you can see at the input there is a transconductor, so no voltage drop (filter.jpeg)
at the output of DUT there is a buffer wihch isolate the filter output from measurement tool, it's a transconductor which loads the the 50 ohm impedance of measurement tool
and there is a reference path to deembbed the effect of the output buffer

please tell me what do you think of that
do i need matching networks, i really dont see the point


PS: this is excerpt from NAUTA thesis

depends on your situation. Not impedance matching will throw away signal power. IF you have a very good signal to noise ratio, and have extra system gain, it is not a big deal. If you have poor signal to noise ratio, your system range will suffer.

Another issue is that filters are designed to work with a specific impedance driving and loading them. If you have a very poor load impedance for your filter, you will end up with various amplitude and phase ripples, and a different bandwidth than designed for. These might be ok for your system, but if it is complex modulation (like 256 QAM and a big bandwidth) it would be a show-stopper.