@tk I saw the same curve on a dual band vertical I foolishly mounted within several wavelengths of aluminum siding. The sawtooth you see looks like the interference pattern from RF reflected back into the dipole, and the phase changes with wavelength
One thing I’m looking at is the railing parallel to the dipole at about 70cm below, which looks like a giant yagi reflector element. Maybe you can test this by temporarily changing the orientation of your dipole. Maybe you can make the railing part of the antenna with a gamma match? Just throwing stuff out there