A direct comparison between in situ and radar measurements of temperature spectra of atmospheric turbulence was performed by Dalaudier et al. <1989>. The widely accepted hypothesis of scattering from isotropic turbulence in the inertial subrange was used for the oblique reconstructions. However, the results were ''puzzling'' because neither the shape, the dynamic range, nor the power level was correctly reproduced. Other investigations were thus needed in order to determine the causes of these disagreements. A new attempt was performed using essentially the same method based on the same model and a more extensive and improved data set. The results, presented here, are now very satisfying for most of measurements. This agreement allows us to understand that some errors and technical problems were at the origin of the discrepancies described by Dalaudier et al. <1989>. From the new results it can be concluded that the method used, based on the isotropic model in the inertial subrange, is adequate to fairly reproduce most of the VHF radar measurements at 15¿ from the zenith. ¿ American Geophysical Union 1996