Whenever you have a mismatch, a length of coax will rotate that mismatch around the smith chart once for each half wavelength of coax. Each time that yellow trace crosses the horizontal line in the smith chart, the purple trace crosses it’s 0° line (may need to select that trace to see that), so having the purple trace on is redundant and doesn’t offer much else here.
Not sure what dielectric that coax has, but assuming a VF about .7, that comes out to about 20’ of coax based on those 13 trips around the chart that I could count. So there is about 20’ of common-mode path, and this time, the VF of that path is not as low because there is not a second conductor, so lets call it .95. The lowest frequency where that common-mode path is a quarter wavelength (to better accept the current), is 6.3MHz, and it also accepts power on odd multiples of it. So the common-mode path also accepts power on 422, 435, 447, 460, 472 etc. Now, this may not correlate perfectly with your SWR trace, but I made some serious assumptions and estimations here to illustrate the point.
For that reason, I would also add a clip-on ferrite to the coax at the antenna end to prevent the coax shield from being part of the antenna. Or use a balun of some type.
