It sounds like the information density of a given syllable is decided subjectively. The subjective measurements are probably not proportional. It makes sense - Imagine if you're tasked with giving a numerical value for how much information is in a sentence. Working through the sentence syllable by syllable and picking an arbitrary 'information score' of each syllable, then adding the information scores together won't get you a particularly good result.

The conclusion of the article (that all languages transmit data at the same rate) is supported by the research, but not proved by it. I suspect that if you read the paper, the researchers wouldn't make it sound so conclusive.