Do you mean, the first returns false and the second returns true?
Ah you're right, thank you for pointing it out!
In the previous version of this comment (where I was still reading it incorrectly) I added a fun fact, that the significand of an IEEE 754 double-precision float is only allocated 52 bits, but the "hidden bit trick" provides an extra bit of precision when the normalized form starts with 1.