Assuming that the 'username', 'BatchID', 'size', 'Date', 'Status' and 'Owner' fields will never have embedded whitespace, then this is perhaps not more 'efficient', but I think more maintainable:

>perl -wMstrict -le
"my $rec = 'user2   1707  amanda test 1 .doc.412   4064   '
         . '11/07/08-15:47   CREMFY   roadtest'
         ;
 print qq{'$rec'};
 ;;
 my ($username, $batchid, $remainder) = split /\s+/, $rec, 3;
 my $filename = reverse((split /\s+/, reverse($remainder), 5)[-1]);
 ;;
 print qq{'$username' '$batchid' '$filename'};
"
'user2   1707  amanda test 1 .doc.412   4064   11/07/08-15:47   CREMFY
+   roadtest'
'user2' '1707' 'amanda test 1 .doc.412'
[download]

Update: Something like this approach would work well if the data fields are variable-width; if they are fixed-width as the example data of the OP suggest, the  unpack approach of Re: read the filename column would be far better.