My take on this.
I would not worry.
Once you have normalized the volume of the 24 bit recording ( increased the volume to closer to 0dB ) there is not much use in keeping the extra 8 bits. Simply dither down to 16 bits.
The fact is that no 24 bit recordings has anything but noise in the lowest several bits. You can check the specs of your box. The really good ones comes up to 110dB S/N, many lower priced ones struggles at 96dB or lower. Theory gives about 6.01 dB plus or minus some change at the total for each bit. So 96dB is 16 bits ( equal to CD quality ) and 110dB would be a few bits more. Theory and practice both confirm that the difference to 99% is only a question about the position of the noise floor. Some purist say they can hear some extra definition, I definitely cannot.
I use the extra bits in 24 bit recordings to allow a larger headroom before clipping. I move the input volume down a bit, say 12 dB, to get that extra headroom for handling increased volume or unexpected things. I move the volume up in software when home before publishing.
The main difference between recorders generally is not the number of bits. Instead it is about the quality of the circuits. I would rather record with a really good 16 bit recorder, say an old-time professional DAT, than one of the lower quality 24 bit machines. In my experience the really large difference is in the analog circuits. Adding a good preamp really helps more than going to 24 bits.
// Gunnar