I'm sorry. I still don't see "taken as a whole" in that original statement of yours.
Moving on...
Actually, 1 out of 3 overall people prefered bass that is different from Harman target. I see 1 out of 3 people as significant.
The qualitative descriptions for each of the three groups has an age correlation as part of the description.
You really should read the actual paper you're talking about instead of drawing rash conclusions from from a third-hand source - a blog post, that misrepresents a presentation, that's summarizing an
AES paper. So let's look at the actual conclusion of the primary, instead of tertiary, source - the AES paper:
In this paper we conducted cluster analysis to identify different segments or classes of listeners based on their headphone sound preferences. The analysis included preference ratings of 130 listeners for 31 different models of AE/OE headphones reported in [1]. The demographics of the listener (age, gender and prior listener training) in each class were examined, as well as the acoustic features of the headphones preferred in each class. A summary of the three classes of listeners is given as follows:
Class 1: “Harman Target Lovers”
They make up the majority of listeners (64%) tested, and prefer neutral sounding headphones equalized to the Harman Target response curve. Membership includes an approximately equal balance of members across gender, age groups, and trained/untrained listeners. The exception is listeners over the age 50 who are more likely to be members of Class 3.
Class 2: “More Bass is Better”
This is the smallest class (15%) of listeners who prefer headphones with 3-6 dB more bass than the Target curve below 300 Hz. Members in this group are predominantly male, and include 30% of the trained listeners in our sample.
Class 3: “ Less Bass is Better”
The second largest class (21%) prefers headphones with 2-4 dB less bass than the Harman Target curve below 100 Hz. Membership is comprised entirely of untrained listeners, and predominantly female and older listeners (50+ years).
Due to there being three (not necessarily independent) demographics quantified in each class, it's simply not possible to come to firm conclusions about any possible correlation between age and preference. For example, it's plausible that older women are least likely to be trained listeners, and the primary causal factor in Class 3 is in fact being untrained, with age and gender being mostly correlative factors. The fact that this class is composed of 100% untrained listeners is also suggestive of this. It's notable too that Class 1 is not only the majority class, but incudes the widest range of demographics, suggesting it's the most generalizable over a wide range of people, as well as the most popular purely on numbers, both of which are important for a target response to be useful. Also notable is that the exception of over 50 years old is the age most people can only hear up to 12kHz, which is breaching frequencies in music that are fairly common, and so reducing bass may compensate for this perceived darkening of spectral tilt when listening to music. So hearing loss may be a hidden causal cofactor (possibly not only age-related hearing loss). Or maybe not. It's impossible to tell from this data. So no, this research does not show there's a big deviation from Harman target preference; it in fact suggests the majority of listeners from the widest range of demographics prefer it. Moving on...