it's important, a) the amount of dynamic range and b) when you bounce down to 16 bit, a dithering process should be taken and dithering 16 bit's to 16 bit's is something to be avoided. i can't say for sure but you may be making the comp and/or the program work a little more harder due to the mismatch (it already does a little more work in 24 bit's when 32 bit vst(i) are being used instead of 32 bit's (it's not a big hit or necessary). the difference between 16 and 24 is significant with 16 being more limiting, but do you, you won't gain anything upconverting if the file is already 16 bit