Yes it makes sense, but no it doesn’t work like that 
Audacity can only record the data that it received. If the audio data gets “converted” in any way before it reaches Audacity, Audacity can do nothing about that.
Audacity works internally in 32-bit float and can convert from 16-bit integer or 24-bit integer to 32-bit float perfectly.
Currently, if you want to record 16 or 24 bit and export as the same with no processing at all (where all exported sample values are exactly the same as Audacity received them), then you need to temporarily disable “dither”. In future this workaround will probably not be necessary.