Thank you all for your answers.
I made some new tests, avoiding white colors this time, and the pictures are still underexposed (as the histograms show it).
The only thing that comforts me is that I get the same result using the builtin flash of the D80 (my only point of comparison up to now), which probably means that my Yongnuo is not faulty (furthermore, it works well in M mode, and, in i-TTL mode, the flash output changes if the scene changes, so it is not stuck at a minimum value either).
http://baptisteautin.com/files/DSC_5504_yn_0_histo.jpg
YN568 i-TTL (f=44mm, M, f/4.5, 1/160, ISO 320, spot)
http://baptisteautin.com/files/DSC_5512_builtin_0_histo.jpg
D80 built-in flash i-TTL (f=44mm, M, f/4.5, 1/160, ISO 320, spot)
So, well, I think I'll keep my Yongnuo... Maybe I'll even get another one, a 560 for example, to add more light, if needed.
And it's true I'm not using a very luminous lens (an AF-S Nikkor 18-70mm f/3.5-4.5G), so I guess I can't expect to get the same result as with a, say, 50mm f/1.8, for example)
Yet I would be curious to see how an SB-900 would perform in the same conditions, in terms of metering, as well as in terms of global power.