If I follow your question correctly you are asking a similar question that I asked when I was still relatively new to DNA testing but have a better understanding of it now.
My challenge with your question is it lacks context for the specific endogamous group as well as it is very 'broad' question of "What is the highest known cM value for a triangulated group of matches that share the same ancestor from the 1700s?" as there is for example going to be significant difference between say a Pennsylvania Mennonite population that still has some in/out migration in "maximum cMs" , or very broad for example Jewish lines, or a say a smaller population isolated to an island or a mountain village and how many generations are moved away.
Then you still will have generational loss of data over the 200-300 years since then even in an endogamous, the age / generation of the testee (i.e. my grandfather matches some at 90cms where I don't even match some of them) , who their testing service was, are the family trees you are referring to even correct or some assumptions being made, and 'some' mutation so it is hard to clearly state.
I have't seen a formula posted that takes into account population size, context, time to approximate it and come up with a 'maximum' cM like you are asking.. but I will share my experience at least with the limited information for a broad question.
I as a general rule I use is that I expect that mixing / noise of elements below 15cMs to not be too useful in the general population because you will get bits and pieces of cMs that do pass through generations from a group through intermixing and on charts they may look continuous but maybe they are not quite so and just appear to form long blocks. Tools like on GEDMATCH can tell you if they are true blocks.
The situation I deal with is a endogamous like situation family in a region intermarried in the small sized communities in a western PA region over the generations back and forth through the 1700s the 1800s before migration enough new population started to mix it up a bit.
I see solid blocks like you do in the 10+ cM range (as mentioned in the post above) where the most common ancestor pair for about a dozen families scattered out from there was born ~1815 and starting having children in the 1830s that most have in common.
The maximum block length for them is about 10cm in Chromosone 1 that they all have and about 3cm in Chromosome 12 which only about half have.. though their aggregate cMs for that born 1815 ancestor pairs is still relatively high between the elders (mixed genders) tested.. (I have 5 who are in there 90s today who share 80+ cMs for that 1815 , but those in there 30/40s aggregate CMs are much lower).. but these clear cut block patterns for my particular group are disappearing after about a 100 years of being away from the area after about 2-4 generations of moving away from the area with those really big blocks anyways for the lines I have research so far.
To my previous comment about clear cut blocks disappearing after about a 100 years.. I also am tracking a couple of families trying to find a common ancestor from a low population that several male lines share Y and we've tracked lines back to 1770s but have no common ancestor and in the aDNA tests they do not show up at all or barely as matches under 20 cMs.
Autosomal DNA tests are generally good for 4-6 generations and with closer nit groups I've been able to extend that to about 10 in a few rare cases.. but 1700s is a bit challenging to say.. as pushing 10 unless a very close nit community of marrying 1st cousins I'd expect there to be some dispersion...
So personally with the limited information provided I think 20cMs seems high for a common ancestor in the 1700s for the limited information shared.. and you would fall into the 'noise' category.. and those nice common blocks you and I have just happen to be preserved over time but I've seen them disappear in a single generation completely... though using a combination of DNA and documents you can fill in the gaps..
The best example I can think of any tool out there right now, and while it is still in the works and constantly being revised is Ancestry's "Origins & Migrations" tool that merges DNA results with location information and establishes 'logical' groups based on it.
I know this does not take specifically provide a definitive answer but I see your question as pretty broad, but understandably so as well, so hopefully my reply helped to get to the root of what you are trying to actually ask.
So "Could it be that a group of matches, on the same segment, with the cM value in the mid 20s, are out-of-range?". My answer is Yes, they definitely could be out of range especially for the 1700s based on the information I posted above.