Just because you have seen them repeatedly in the models, it doesn't make them quality polls. Look at the 538 rankings for these polls...
https://projects.fivethirtyeight.com/pollster-ratings/
7 polls from Insider Advantage - Rank 97 - 2/3 stars. Transparency 3.3/10.
7 polls from Fabrizio/McLauglin - Rank 136 - 1.7/3 stars. Transparency 3.3/10.
6 polls from OnMessage - Rank 238 - 1.1/3 stars. Transparency 1.7/10.
3 polls from Trafalgar - Rank 279 - 0.6/3 stars. Transparency 1.1/10.
Note the transparency rating. It means that quite often these polls won't publish the crosstabs, how they sampled or data on how the poll was conducted.
Best example of a ridiculous GOP poll is the recent "American Greatness" poll, which is included in the 538 model still. The ran a PA poll and largely excluded Philadelphia!!!
https://www.newsweek.com/pennsylvania-poll-tipp-trump-harris-likely-voters-philadelphia-1967590
Yes, these polls will not be weighted as highly as ones from respected sources who work with major Universities, but if you throw in enough junk polls into the model, they will start to make a difference. Especially as the polls recency gets a high weighting. And those from swing states what get an every higher weighting into shaping the national model. And we are talking here a huge number - 27 of the most recent 60 polls.
I dont think its being overhyped.