I'm trying to build a small magnetic loop for 5W on at least 40m - 20m following ideas of G4ILO's wonder loop. For this, I'm using coax with a ~2mm copper core and ~1mm spacing to the braid (see 1st image below), with braid and core connected on the ends. I use 50cm of this cable for the coupling loop and 2.90m for the outer loop (which is not based on any calculation but on the length I had available). I use a 350pF variable capacitor with ~540° rotation. The antenna is placed ~12m above the ground.
I first placed both loops as circles (see 2nd image below), pretty much as the image on G4ILO's page. I could peak the noise level on 40m and dip it but not with a satisfying SWR at all. I couldn't peak the noise level on 20m and didn't try to dip it.
With some experimenting, I found out that RX improved significantly when the loops touched each other as much as possible, so I experimented with an egg-shape coupling loop inside a circular outer loop. This gave somewhat better results.
I then kept the coupling loop somewhat circular and made the outer loop pear-shape so that it fits around the coupling loop (see 3rd picture). This allows me to peak noise level on 40m and dip it with near-1:1 SWR. The dip is rather sharp and I need to retune every ~10kHz. I can hear a fair number of stations but don't get any results on the reverse beacon network / webSDRs when I CQ myself. Normally, I get several spots on the RBN (with a 20m dipole, up ~5m).
On 20m, I cannot peak noise level but I can dip to 1:1 SWR (perhaps there is so much noise on that band that the radio's AGC levels it, making it impossible to peak noise level?). The dip is much wider, I don't need to retune on 14.000 - 14.060 at all. However, also on this band no results from the RBN yet.
My question is twofold:
I know that a 1:1 SWR does not have to be ideal (a dummy load also has 1:1 SWR). Is there a fundamental problem with my latest construction that allows me to get a 1:1 SWR but little radiation?
What can I try to improve the current construction?