This has to be launched by Ariane - it's constraints are tightly bound to what that platform can do, and it's fairing size. Ariane is optimized for GEO insertion, while Falcon is optimized for LEO orbits. You could have used a Falcon, but a payload like this was actually built around the rocket's capability, and this was designed prior to Falcon being a thing.

All that said it's worth nothing that SpaceX's flight success rate is 98.5 (135/137), while Ariane V's is 95.5 percent (106/111).

The really gobsmaking thing about that is that this is that SpaceX's rate is over 11 years, while Ariane's is over 25 years.

It's time to stop thinking of SpaceX as the plucky, untrustworthy startup.

In the future space telescopes like this really need to be built in LEO, and then boosted to Lagrange points. The number of failure modes beyond the typical rocket / stage / fairing, secondary burns that the folding mechanism and the lack of a ability to test a ton of new technology in zero-g orbit makes this far more likely to fail then anyone is comfortable with, given the overall cost to this.

Important to note that the agreement was made in 2007, so reliability as a factor would be reliability as assessed in 2007, not 2021.

I think that in 2021, Falcon 9 ‘s track record arguably suggests it is more reliable than Ariane 5, but it doesn’t matter because the Falcon fairing is too small for JWST.

Falcon 9 has a better track record than Ariane 5 at this point, but SpaceX wasn't really a thing at the time these contracts were signed.