Irenaeus (Adv Haer III.22.3) seems to have known a text of Luke with 72 generations: "Wherefore Luke points out that the pedigree which traces the generation of our Lord back to Adam contains seventy-two generations, connecting the end with the beginning, and implying that it is He who has summed up in Himself all nations dispersed from Adam downwards, and all languages and generations of men, together with Adam himself."
But I can't locate a manuscript reflecting that number, although there are lots of possibilities: e.g. Bezae has 65 names; 1071 has 73 names; Vaticanus has 76 names; Sinaiticus has 77 names; Alexandrinus has 74 names (W and 579 omit the whole thing).
NA27 prints a text with 77 names, this has the advantage of neatness (significant sevens all over the place; cf. Bauckham in Jude and the Relatives of Jesus); but the disadvantage that this neatness is pretty much imposed on the textual evidence (esp. at 3.33 reading Aminadab, Admin, Arni). Is that a reasonable approach? Did Irenaeus make the whole 72-thing up or did he simply miscount? It is interesting that he draws theological significance from a reading for which we have currently no evidence. Should we accept that the textual evidence is too uncertain to allow far-reaching theological deductions from the ennumeration (M.D. Johnson, Purpose of Biblical Genealogies)? Or should we allow the theological neatness to help us determine the original text?