I can only think that it comes down to being able to predict
relationships. Using the longest segment would not allow companies to
predict relationships as they do now with total cM's shared. Or at
least not as easy.
The recombination process that takes place with autosomal DNA is random.
It would be much harder, if at all possible, for companies to predict
how many crossovers would exist in each chromosome and where they would
be. Also, chromosomes vary in size. Without this, one cannot determine a
standard segment length to associate with a particular relationship.
The use of total cM's shared is a very good predictor up to 2nd cousins.
Beyond that, the variation observed in total cM's shared as a percentage
of the expected average cM's shared becomes more pronounced and this
makes it more difficult to predict relationships. So i think you did a
valid exercise by also sorting by the longest segment. Still i guess
total cM's shared is the best way companies can, with existing
technology and knowledge, predict relationships.
Rio de Janeiro, Brazil