Same as Fig. 6, but here the h(λ) variations are binned over 200 pixels (~2 Å) for clarity. Here the altitude variations are, following Eq. (14), extracted from the eclipse observations of 20h44 UT with either the 02h56 UT (solid line) or the 00h12 UT (dotted line) full Moon ones. Without the binning the difference was not visible. Here we can note that both possible perturbations related to either the BERV shift as well as the air mass correction are indeed negligible (see text). On the contrary some broadband variations could be seen, probably linked to albedo variations of the different lunar regions where the fibers were pointed (see Table 1). In the blue region (4000 to 5000 Å), the fringes are due to an imperfect correction of the SOPHIE orders and not to classical CCD interferential fringes (see text).