You've touched on what is probably the real issue, but without further measurements, we cannot know for sure. The author said (in the comments) that they used

You've touched on what is probably the real issue, but without further measurements, we cannot know for sure. The author said (in the comments) that they used an antenna as the load. With a minimum SWR of 2:1, it's obviously not a very good match. If the match is bad, the length of the cable could have a big effect on the measured power. Other comments suggested using a resistive dummy load instead of an antenna. THAT is the way to go if you are measuring cable insertion loss. The two cables with the poor insertion loss were 3 and 10 meters in length. None of the good cables were of these lengths.