It also feels like BS to me. Multiplying the information density of the average syllable with the average number of syllables spoken per second should give the amount of information conveyed per second. The claim is that this is constant, but the table given does not support this.
I saw another similar but broader study a while back that singled out Japanese as the (low) outlier, and a claim that speakers of Japanese make up the shortcoming via facial expressions.
Let's make that calculation:
That would make English carry 47% more information per second than Japanese.