This happens to me a lot. Some of it is due to certain family lines I have which tended to intermarry -- others just seem to happen. On the whole I find the prediction that any of the companies (I have data at Ancestry, FTDNA, and GEDMatch) are way more close than they are in actuality.

Roberta Estes has some blog posts on this phenomena:
http://dna-explained.com/2013/10/21/why-are-my-predicted-cou...
http://dna-explained.com/2014/02/19/generational-inheritance...

I'm sure some of the other bloggers have delved into this as well.

The (unscientific but descriptive) term I have seem some use is "sticky DNA". As in segments that want to stick together and get passed along whole rather than being broken up -- which is what the you inherit 50% from each parent rule predicts would happen with the segment (i.e. you'd only get 50% of that segment length).

There always could be a close connection. How well mapped out are both trees?

Elizabeth