Ah yes, I remember Greg writing something about the tap length (=number of samples) being the same for all sample rates, which is unnecessary and should be changed in a future release.
Tap length should be half for 44.1k compared to 88.2 KHz to cover the same length of time, which is the relevant aspect of convolution file size.
So the lower res convolution files has alot of unneeded zeros in it.
I use Filters created using the FocusFidelty Filter Designer, which are 33k Taps for 44.1 kHz and accordingly increased tap length for higher sample rates.