I think they meant: the quadratic lower bound is the highest yet proved -- not that it is "better" than Knuth's unproved conjecture.
It is worse in the sense that it tells us less about the actual complexity class the problem is in, so I guess it depends on what you interpret "better" to mean in this context.
lower bound like you are talking about (the function always takes longer than this to run for an input of size n)
lower bound across all inputs of size n (the function never runs faster than this, given any possible input of size n)
The two are different
EDIT: adding a link to the wikipedia article defining Big-O and Big-Omega: http://en.wikipedia.org/wiki/Big_O_notation
>It is not contradictory however, to say that the worst-case
>running time of insertion sort is [Big Omega](n^2),
>since there exists an input that causes the algorithm
>to take [Big Omega](n^2) time.
>Introduction to Algorithms, 2nd Edition, p 46It's worse in the sense that the problem takes more time to solve than if someone had found an algorithm in O(n^2).
EDIT: spelling
"Knuth had previously proved an exponential upper bound on the number of Topswops steps, and conjectured that one might also prove a matching lower bound. What Dr. Hal Sudborough and Dr. Linda Morales did, however, was to prove a lower bound that is much better than that proposed in Knuth’s conjecture..."
However, the paper cited says “A Quadratic Lower Bound for Topswops”, so the lower bound proved is much worse than the one conjectured by Knuth.