The availability of high-quality archaic human genomes has revealed a complex admixture history of modern humans. However, it is still unclear to which extent neglecting past admixture events with archaics has biased our view of the evolutionary relationship (e.g divergence times) between modern humans (MH). To address this question, we first performed computer simulations under various scenarios including archaic admixture and show that if two modern human populations are differentially admixed with an archaic population their divergence time is strongly overestimated. We then quantified this bias by using a data set including high coverage genomes of Aboriginal Australian (AA), Eurasians, African and archaic humans. A likelihood analysis of the multidimensional site-frequency spectrum reveals that neglecting archaic contribution into MH leads to a divergence time between AA and Eurasians that is considerably older (~20 ky) than what is estimated under a much better supported scenario that explicitly accounts for differential archaic introgression into these populations. Finally, under our best scenario, we estimate that admixture from Denisovans into MH occurred as recently as 31-55 kya. We conclude with a discussion on the implications of our admixture times for the population history of our species.