Of course, as many people here are saying, the analysis is based on written text and reading out loud, not spoken conversations, which would probably require a different analysis.
I think it's more natural to measure the spoken expression rather than the written expression, specially for Japanese as all the major writing forms[1] were borrowings made to fit the spoken language.
[1] I guess that's pretty generic as all writing forms are made to fit. What I mean is that Kanji characters were fitted upon Japanese language --which had not written form. Basically, the way I understand it, they took a word like Mountain (Yama) and then would ask, how do you write mountain in Chinese(山)/shan/ and then would say (山) = /yama/. Now let's say there was another word that had the sound /yama/ but had nothing to do with mountain, they would still use (山) to write it. so rather than being an ideograph, they became phonographs.
Kanji were fitted upon Japanese as much as they were upon Chinese. Kanji is a writing system that doesn't depend on phonetics in any way, the correspondence is one-way only.
However, the study is still quite interesting. Besides the syllabic analysis and derived information density, they also analyzed recorded speech for the time necessary to convey equivalent semantic content, and while this metric diverges somewhat from the others, Japanese still comes out at the bottom of the group. It's worth pointing out that if this hypothesis is correct, it's not at all in conflict with anecdotal observations about the high context in Japanese conversation. If anything, one would expect context to take over when formal information density is low, much the same way that tonal languages often have comparatively low phonemic inventories.
I do natural language processing in Japanese and typically texts in Japanese say more in less space than English, by far. Both in spoken form and in written form. They say about as much as in Chinese, when both texts are written by native speakers. Now, you cannot compare texts 1-to-1 because when texts are translated, they are remarkably less information-dense. There is a whole lot lost in translation and "not-assumed" when the original writer didn't have a native Japanese context awareness. Reversely, when the text is translated to Japanese it will often be less information dense but not as much. Partly, because the translator will find a balance between not adding information (a basic translation principle) and not sounding unnatural, as this would also take away the attention from the reader. Both cannot be achieved to a high standard between languages as different as, for example, Japanese and English. In the tests I do, I can distinguish with remarkable accuracy when the text was originally written in Japanese or not, and they are all native texts and translations.
Japanese is not just a language, it's also a whole culture and civilisation to an extent most other languages are not, except maybe Korean (who are some sort of bizarro Japanese to be honest... just don't tell either them or the Japanese). Chinese language is shared by remarkably different cultures and ethnics, they don't have the same level of common context. European languages have nowhere as much context.
In Japanese, you'd have a conversation going like this:
A "So, should I leave the umbrella by the door or can I put it to dry open in the aisle?" B "un" (colloquial Japanese "yes")
And, given features that are totally lost or implicit, they'd now exactly what they mean. They'd answer negative, double-negative, even triple-negative questions with monosyllables and know exactly what they meant without any confusion or hesitation.
For me, the benchmark is seeing how much is lost in translations and subtitles in films. Japanese ones subtitled in English lose a massive amount of detail even in the most common conversations. The converse is nowhere close as remarkable. They simply have a much more complex set of rules for social interaction. A lot of what they say is customary but also untranslatable. A lot of what they don't say goes completely unnoticed by a foreigner, because the omission of these customary expressions also carries a lot of meaning. In short, it's not possible to compare the expressiveness of such a different culture, and the way we are even trying to measure it just shows how much we're "not getting it".
For example, would not "strengths" be considered one syllable?
I don't know about Japanese, but Mandarin has a linguistic syllabic structure that would not allow for such long syllables.
Actually I'd say it's one of the most efficient languages I know in terms of what needs saying and how quickly they get their point across.
Now, drawing any comparisons is extremely hard because context doesn't translate, so by definition you cannot create a benchmark. For instance, they convey far more information about social nuance and politeness, to levels that would need whole paragraphs per sentence to explain.
Also, syllables are a horrible benchmark for Japanese since all their syllables are extremely short and quick to pronounce. Their encoding, so to speak, consists of a smaller unit size and a bigger number of units. They have a compression system for text, which is the mixture of kanji and kana.
This basically goes to show how unique Japanese is in many respects and how differently they communicate.