I used a Garmin GPS 64 to record altitudes during a recent hike along the the Vidourle river-mill circuit and compared the recordings to the theoretical altitudes obtained by generating the GPX with an online tool using the IGN basemaps. Here below is a comparison chart.
First, notice there is a growing lag between the peaks on the theoretical profile (green line) and the recording (red line) with the peaks on the recording occuring later then the peaks on the theoretical profile. This is normal, because we did not stick exactly to the theoretical track. At Montredon, we wandered around the site which added some distance to the recording that does not appear in the theoretical profile. (Note: We also climbed a bit higher than planned.)
However, there is still something strange going on. The recorded Garmin altitude, starting from around 3 km in, starts tracking progressively lower than the theoretical altitude. The Garmin data stabilises between 10 and 20 m altitude while in reality, no point on this circuit is below 20-25 m so the Garmin data is incorrect. Is this because the differences are within a margin of error for altitude measurements with a handheld device ? Maybe but if so, why is the tracking much closer for the first 3 km ?