I found some versions of diskspan very temperamental, I use the following hashed version for the best results, make sure you're only specifying the first archive in the diskspan converted output, and to not rename the outputs to anything other than the initial output. For my archives I use "l33t-XXX-1.bin" as the output, diskspan will then use that numbering for subsequent split outputs.
Code:
MD5: 255783423af3f6e9513d6600685d51c8
As it's a CLS plugin you also won't need to specify anything in the arc.ini for decomp, in your method string, you just need to specify the size of the split archives, i.e. diskspan:19.98gb
The record.ini entry would look something like this if you're struggling.
Code:
[Record24]
Type=Freearc_Original
Source={src}\l33t-XXX-1.bin
Output={app}\
Disk=24
Component=2