OK, I realize this is not exactly what you asked for.

However, davido has written a very cool module called PerlMonks::Mechanized (beta), which IMHO, deserves more "press". bobf has extended it with a method called user_nodes. The code below assumes that you have added bobf's user_nodes method to davido's module, and will print the node ID, document type, and node title for each "original" user node (in the sense of "original" which you have elucidated in your OP and its replies).

Please be kind with any comments on my code; I am still learning. :-)

#!/usr/bin/perl -w
# GetUserRootNodes.pl
# usage: perl GetUserRootNodes.pl username password 

use strict;
use warnings;

use PerlMonks::Mechanized;

my $user = shift(@ARGV);
my $passwd = shift(@ARGV);

my $pm_obj = PerlMonks::Mechanized->new($user,$passwd);

my $data = $pm_obj->user_nodes;
my $nodeids = $data->{NODE};
while (my($nodeid)=each(%$nodeids)) {
    my $info = $pm_obj->node_info( $nodeid );
    if (not (exists( $info->[0]{root_node} ))) {
        print $info->[0]{node_id} . "\t";
        print $info->[0]{nodetype} . "\t\t";
        print $info->[0]{content} . "\n";
    }
}
[download]

Sample Output:

491054  monkdiscuss             Free Nodelet Hack: No more scrolling t
+o find the "comment on this node" link
442602  user            planetscape
454613  perlquestion            Help with a Regex
471590  perltutorial            Not Exactly a Hash Tutorial
475126  monkdiscuss             Free Nodelet Hack: Check or Clear all 
+"Delete" Message Checkboxes in Your Message Inbox
[download]