Curious GPS behaviour walking both sides of a stream

I’ve been mapping a meandering stream (one recently created by a stream restoration project) by walking both sides of it. The method I use is to always keep the stream on my left and hold the GPS over the edge of the water. A lot of this stream is in a wood and I don’t expect good GPS accuracy, but I noticed a funny thing on several days work.

I expected that most of the time the outgoing GPS trace would be to the right (relative to my initial direction of walking) of the returning GPS trace, because the stream obviously has positive width. But I find that the opposite is the case.

In the example below, I started top left and walked the cyan-coloured route, keeping the stream on my left. I then crossed over the stream and walked the magenta-coloured route, again keeping the stream on my left. (Ignore the small section in the middle where I took a different route when returning.) In this example, the cyan trace should be SW of the magenta, but it’s clearly more to the NE.

My GPS unit is a satmap ACTIVE 10, which is doing quite well under the trees IMHO. But can anyone explain why there is this systematic displacement of the return trace relative to the outgoing one? (Of course, we don’t know which, if any, is more “correct”.)

This could possibly be explained if the traces were taken at different times of day - the tracking satellites may appear at different points in the sky.