It's possible that the data-set was even designed to make this an issue.

Could be. It is a 'simple' problem that is deceptively hard to accomplish given the constraints. Time in this case.

Howver, I picked another of the easy tasks at random -- namely: TIDRICE -- and submitted the following:

#! perl -slw
use strict;

## substitute <DATA> for <> in 3 places to test 

for( 1 .. <> ) {
    my %h;
    for( 1 .. <> ) {
        $h{ $_->[0] } = $_->[1] for [ split ' ', <> ];
    }
    my $score = 0;
    /\+/ and ++$score or --$score for values %h;
    print $score;
}
__DATA__
3
4
tilak +
tilak +
tilak -
tilak +
3
ratna +
shashi -
ratna -
3
bhavani -
bhavani +
bhavani -
[download]

which produces their exact required output from their sample input on my machine; but it is rejected as producing the wrong output.

So, I'm not sure if that means I missed some subtly of the spec. or their testing is flawed.

I haven't reached a conclusion about the site yet. Their UI certainly has some annoyances -- but hell, I've hung around here for 10+ years and this place is far from perfect :)

And one can't but wonder if perlsufi pulled off a very subtle spam with this thread :)