After trial and error and reading the
http://www.sleuthkit.org/sleuthkit/docs/skins_fat.txt, I realize maybe the route to map the LBA to inode nubmer for vfat partition is not right. As both istat and "blkls -l" use sector, instead of cluster, there maybe no need to do such translation to locate the possible associated file(s).
The MHDD had remapped LBA from 17518937 to 17518941, and it falls to the first partition of the disk according to "fdisk -lu." I then dd'ed the whole partition to hdc1.img. The dd process didn't give me any error, because the bad sectors were addressed by the remap of the MHDD.
[root@test ~]# blkls -e -l -f fat32 hdc1.img $((17518937-63))-$((17518941-63))
class|host|image|first_time|unit
blkls|test.yarbs.cn||1231336092|Sector
addr|alloc
17518874|f
17518875|f
17518876|f
17518877|f
17518878|f
The above means they're all unallocated. This is in accordance with the "fsstat -f fat32 hdc1.img," as 17518874-17518878 is not listed in the fat contents.