Talking to myself...

Did a bit more modeling to get rid of the segmentation check warning
about the coupling loop having segments too small on 40M. I had to
change down to an octal loop for the coupling loop to give me wiggle
room on the loop size.

I found that playing with coupling loop diameter and placement distance
from the bottom loop interacted and changed the coupling loop impedence
presented to the source. In general lowering the loop will make it more
capacitive and lower the resistive portion. Making the loop smaller
raises the resistive portion and raises the capacitance as well. So I
moved it up until the reactance got pretty small then fiddled with the
size until the resistance approached 50ohms then moved it up or down to
again reduce the reactance. Lots of fiddling. I suspect that indicates
a similar level of sensitivity to size and placement in the real world
as well.

After playing a bit I got the SWR down to 1.3 or so.

Retuning the capacitor back to the 23pf and retesting at 20M shows a
reasonable match (1.2) there as well.

So it seems that a likely approach for tuning the match loop would be
to do it on 40M then test at 20M where it appears to be a bit less
critical.