That might be because those characters didn't have players attached who could make decisions for them e.g. the player of the low-stat PC might intentionally go for a more cautious approach, or do things differently e.g. go missile-heavy instead of melee.
But that goes back to how you create characters. If I envision Sir McStabsalot, stalwart gnomish order of the owl paladin, it's because I want someone who wades into battle and doesn't give a hoot. If I roll poorly and have to hide in the back, that concept has flown out the window.
Also, did these cage matches involve a whole party against the monsters or just the single character? There's a huge difference.
If you're saying that the guy who would normally die first is shored up by more capable PCs, sure. It's not about whether or not they could survive, it's how much they can contribute. I'd have to have a simulator of a group that happened to all roll poor stats vs a group that happened to all roll good stats to do a comparison. Because the point is, how much can that individual contribute mechanically to the group.
Obviously D&D is a team game. If you and yours find it more enjoyable to roll stats, coolio. I just don't see any advantage to it and plenty of downsides for me.
In our 1e-variant games at least, I ran some numbers once that showed me starting stat averages really aren't a significant predictor of a character's career-length expectancy.
The guy that hides in the back and runs away at the first sign of trouble may survive longer. Doesn't mean I want to play that PC.