I'd be suspect of the cal. If the cables and jig called at 20M the loss should
be zero and the SWR also perfect (cal absorbs errors).
The S11 should be a good match inside the pass band, its a symmetrical
For S21 its too good. I can imagine and model a better filter loss but
only with perfect caps and coils. See ELSIE.
1.25Db is best I could ever do for double tuned and that was with
3/8th" ceramic slug tuned inductor ( with enough turns to keep
core barely inserted) with silver mica caps critically coupled.
That was verified with both VNA (agilent 8573) NanoVNA, and
S11/S21 setup on Rigol DSA815T. I tend to do that want I get
numbers close to the book. I have found the NanoVNA I have
to be a bit more optimistic at HF by variable amounts but under
.1db (good enough).
Please reply on list so we can share.
No private email, it goes to a bit bucket due address harvesting