Don't be so sure it's a mistake on your side. The code that's inside the Polargraph contains a very simple mathematical model of a machine and a hanging pen. "All models are wrong, but some of them are useful", and this model is ok most of the time, but it never exactly reflects the physical reality.
The attached pic shows the addressable space in the blue square. The model assumes that the cords always hang from the top corners of that addressable space, and so the top edge of the triangle formed by the hanging cords is always a) fixed width and b) fixed position.
Of course, it is _never_ actually in that position, though it approximates it a lot of the time. As the pen moves up the board, the actual hanging point rises above and away from the modelled point. This adds some creeping distortion towards the top of the drawing, where the plotted points are higher than they should be.
So I suppose the weird this is that your distortion is in the opposite direction... The pen is actually lower than the model thinks it is, not higher. I've been scratching my head over the weekend to think of a way to model or visualise this.
