I suggest that what we have here is largely a case of selection bias.
The progressive-matrix questions asked here on Puzzling.SE are not generally representative of the full range of progressive matrix questions out there, for a number of reasons. For one, not all progressive-matrix tests out there permit their content to be shared more widely. However, the more pertinent reason is that, in general, people only post progressive matrix questions here on Puzzling.SE when they find them so difficult that they are unable to work out the answer themselves and wish to seek the opinions of a community whose members may be able and inclined to help.
In other words, the easier progressive matrix questions 'out there' do not get posted here by visitors to the site, because people are not seeking help with those - they have successfully worked out the answer or understood the logic behind them; they need no further help. The ones that do get asked are the more difficult ones which require further assistance or explanation, and hence the pool of progressive matrix questions on this site will always be dominated by these trickier - sometimes impossible - ones!
Note that finding a particular question difficult does not necessarily bear negatively on the person attempting to solve it - many of these questions are badly constructed, depend on weak logical reasoning, or arguably have multiple potential solutions that work. This is a known criticism of this puzzle type and partly explains why these questions often tend to be poorly received in this community of puzzle setters and solvers - which might also explain why none of the site's regular users make a habit of posting the 'easier' progressive matrix puzzles here.
To summarise, it seems to me that we have:
People visiting the site to post the harder progressive matrix questions that they have found trickiest and wish to receive help on.
Little interest or appetite within the community to post the easier ones as a matter of routine.
Together, this means the harder questions tend to get posted, and the easier ones do not. The resulting collective output is not a representative sample of the tests that exist out there.