I agree this seems very odd (and I came to the comments hoping for an answer). edit: Since the conclusion doesn't mention anything about the list being faster

I agree this seems very odd (and I came to the comments hoping for an answer). edit: Since the conclusion doesn't mention anything about the list being faster than the array it seems possible that it is just a typo. Also the paper mentions (in section 2) that: Occasionally when a recorded benchmark time seemed to be implausible, (perhaps caused by other activity on the computer system), or we re-tested the data point for other reasons and found a significant variation, we indi- cate the repeated time as well. which seems to imply that the timing experiments were not usually repeated.