ni-6225: bug in ni_pcimio.c ?

I'm using a NI-6225 board for thermocouple measurements and am wondering
whether anybody else has seen this. 

The NI-6225 has 40 differential inputs, which are arranged in blocks of
8 input pairs (AI0-7, AI16-AI23, AI32-39, AI48-55 and AI64-71). 
The first four blocks work correctly. The fifth block, however, outputs
the same as the first block (i.e. AI0-7), and signals connected to
AI64-71 are ignored. 

I just recompiled comedi from the current CVS, but still get the same
symptoms (using kernel 2.6.20, gcc 4.1.2).

To troubleshoot, I did a differential measurement on AI7 (connecting it
to +5V) and used one of the demos from comedilib: I issued

$ ./inpn -d

and got ~4.8 volts on channels 7, 15, 71 and 79, whereas I should only
see it on channels 7 and 15.

Thanks!

-Ingmar Schoegl-