Joining the choir here, I never accept rolled stats in my SAGA games, and yes, the fact that they included it in the core book is a legacy thing. Perhaps so many people have grown so accustomed to rolling stats that they might not consider any other option, even if it objectively leads to worse games.
Rolling stats would be okay in A) a game in which stats don't actually matter that much (like the very first versions of D&D) or B) a game where the life expectancy of characters is very short (like Call of Cthulhu). SAGA is neither, so I fail to see the fun in a long campaign where some players get severely punished and others plentily rewarded for something that they rolled during the first session.
(I don't actually like rolled HP, either, and this is pretty much my only complaint in SAGA. I've houseruled it so that if a PC rolls less than the average of the die rounded up, they get the average of the die rounded up. If you roll more than that, you keep your roll. This keeps a random element in the roll but softens the impact of a bad roll. I balance this slight increase of power for the PCs by not using Destiny Points.)
There is one exception though, where we went with rolled stats and it turned out fine. That is playing with young kids. When they created characters, they were too young to use and optimize with the point buy system, so I let them roll for stats (which they liked very much). As it happened, both rolled better than average results, and one rolled slightly better than the other. But because of their age, they don't percieve game balance issues that well - the focus of the game is in other things. They've never complained about it, and it hasn't been a problem for me as a GM either. But come to think of it now, if one would have rolled really well and the other really poorly, they WOULD have noticed it and it would have turned sour. Ah, well. Guess it's not a good method for kids, either (the standard array might have been the best choice for them).
So... Just Say No!