No, I understood just fine. What I don't understand is the confusion over 'transferring' files. Again, if you needed to copy a file from the one machine to another, how would you do it? That's what you would do here - no magic, no special stage involved, just a simple file transfer: ftp, scp, whatever mechanism you prefer.
You asked what you should use to
load 10 million records, I suggested 'sqlldr'. Sqlldr does not
create data files nor did I imply that it did. DataStage creates both the file
to be loaded (the .dat file) and the file that tells sqlldr
how to load it (the .ctl file) and then you use sqlldr to do the actual work. Your 'batch job' or script can do the file transfer (again,
not needed) and the invocation of the sqlldr command.
And people really don't need to add 'Please correct me if I'm wrong' to their posts... don't worry, someone will.