Steve speaks the truth (not that anyone accused him of anything else!).
When you open a 24 bit file as 32 bit Floating Point, the original file is unchanged. the extra 8 bits are just a mantissa for scaling. There's no conversion loss and the file you start with remains at it was--you just get the headroom advantage of a floating point system while working.
As for the use of Isotope in a two stage conversion, I just learned something (not that I often have to do a sample rate conversion).