Ascent and descent figures should be used with care. They are generated from Anquet's software, using a 50 metre square grid. This means that if the path is in an area with rapid altitude changes (i.e. closely-spaced contours), then the altitude could vary rapidly, giving a
Therefore a profile of a walk along a flat stretch of canal can appear to have gradients - the points on the grid appear on either side of the canal, and are interpolated from that. It also, unfortunately, occurs on high cliffs - unless the path is 50 metres or so from the edge.
This is therefore a limitation of both data and software, and one that can not be fixed without either a great deal more work or rewalking all of the walks with my GPS' altimeter on - and even then it would need a great deal of parsing.