Strange very large disk partitions behaviour !!!

Hi kernel developers,

I have a RAID disk enclosure that contains 16 IDE disks that is
seen via its SCSI interface as a very large disk sliced in 3 QUASI IDENTICAL partitions, here is the fdisk information:

[root@nfs00 root]# fdisk /dev/sdb

Disk /dev/sdb: 2199.0 GB, 2199014866944 bytes
64 heads, 32 sectors/track, 2097144 cylinders
Units = cylinders of 2048 * 512 = 1048576 bytes

Device Boot Start End Blocks Id System

/dev/sdb1             1        699048 715825136   83  Linux
/dev/sdb2        699049   1398096 715825152   83  Linux
/dev/sdb3       1398097   2097144 715825152   83  Linux

So I formated each partition ext3 ( that went ok) and monted the beasts:

[root@nfs00 root]# df
/dev/scsi/host2/bus0/target0/lun0/part1

                    1008G   33M  957G   1% /mnt/p1
/dev/scsi/host2/bus0/target0/lun0/part2
                     672G   33M  638G   1% /mnt/p2
/dev/scsi/host2/bus0/target0/lun0/part3
                     672G   33M  638G   1% /mnt/p3

[root@nfs00 root]# df -k
/dev/scsi/host2/bus0/target0/lun0/part1

                    1056887972     32828 1003168364   1% /mnt/p1
/dev/scsi/host2/bus0/target0/lun0/part2
                    704592112     32828 668768028   1% /mnt/p2
/dev/scsi/host2/bus0/target0/lun0/part3
                    704592112     32828 668768028   1% /mnt/p3

Now I'm really confused, the LARGER partitons are showed SMALLER
then the smaller partitions and why that enormous difference, what I'm doing wrong, I've fsck each partiton and all seem to be OK, I'm afraid of an integer overflow or such thing that can blew the whole storage or I'm too tired and overlooking something.
Please help me while I'm still having some hair left TIA.

Best regards,

Mircea Ciocan

P.S. Kernel is 2.4.21, glibc-2.3.2, any othe info on request.