I don't think the point of the story was to make a sort of scientific accurate comparison. It was more of a fun anecdote of how our assumptions can be turned on their head.

There is any way quite a long list of stories regarding the experience various researchers have with Julia, seeing major speed improvements.

Not to mention that a lot of approaches to speeding up your code in Julia would be prohibitively expensive and difficult to achieve in Fortran because the language is simply too inflexible.

E.g. you could a complex algorithm and train it in a deep learning network on a GPU or tensor proceessing unit using Julia. You could do that in a couple of days in Julia. The same would likely take months of work in Fortran.