Yes. Altering 8 bits of information from the 24 bit file in order to do a so called blind test is a flawed test IMO. One has to assume all 8 bits of that truncated info is not audible for it to be a valid test. Not going to waste my time with that. In the Harley article I linked earlier, Bob Stuart from Meridian said comparing 16 bit masters to 24 bit was like turning on a light switch. Not different mastering, just different bits. Most subjective writers agree with Stuart. Of course, this thread is really about comparing low bit rate 320 to 16 bit. Again, one has to assume the roughly 75% of data removed from the 320 is not audible to honestly say they are the same.
But like I said earlier, I do think 320 sounds surprisingly good considering. 256k is clearly audibly worse than lossless in my opinion.
You're not changing anything in one of the files.
You're just making a copy of the 24 bit, converting it to 16 bit, and then converting it back to 24. Compare that to the unconverted one.
The removal of the bits on one of the files is the entire point... that's how you get it to be a 16 bit file... I really don't know what the problem is here.