The isopycnicity hypothesis states that the lithospheric mantle of ancient platforms has a unique composition such that high density due to low lithosphere temperature is nearly compensated by low-density composition of old cratonic mantle. This hypothesis is supported by petrological studies of mantle xenoliths hosted in kimberlite magmas. However, the representativeness of the kimberlite sampling may be questioned, given that any type of magmatism is atypical for stable regions. We use EGM2008 gravity data to examine the density structure of the Siberian lithospheric mantle, which we compare with independent constraints based on free-board analysis. We find that in the Siberian craton, geochemically studied kimberlite-hosted xenoliths sample exclusively those parts of the mantle where the isopycnic condition is satisfied, while the pristine lithospheric mantle, which has not been affected by magmatism, has a significantly lower density than required by isopycnicity. This discovery allows us to conclude that our knowledge on the composition of cratonic mantle is incomplete and that it is biased by kimberlite sampling which provides a deceptive basis for the isopycnicity hypothesis.