I noticed that the beginning times for each like Team-City pairing was the same. I didn't know if this was coincidence or that they would be the same in every case. This solution doesn't assume that.

#!/usr/bin/perl
use strict;
use warnings;
use Time::Piece; # core since Perl v5.9.5

my @data;

open my $fh, "<", 'o33.txt' or die $!;
while (<$fh>) {
    push @data, [ split /,/ ];
    my ($start, $end) = map Time::Piece->strptime($_, "%Y-%m-%d %H:%M:
+%S"),
                        map s/\.\d\d\d$//r,
                        @{ $data[-1] }[2,3];
                        
    push @{ $data[-1] }, $end - $start; # piggyback elapsed seconds $d
+ata[r][c]
}
close $fh or die $!;

my %seen;
my @lines = grep {!$seen{ "@$_[1,4]" }++}
            sort {$a->[1] cmp $b->[1] || $b->[-1] <=> $a->[-1]}
            @data;

for my $rec (@lines) {
    printf "%s,%s,%0.2f hours\n", @$rec[1,4], $rec->[-1]/3600;
}

__END__
*** input (o33.txt)
1,BEAR,2011-11-21 08:49:16.000,2011-11-21 08:53:13.910,San Francisco,,
+,
2,BEAR,2011-11-21 08:49:16.000,2011-11-21 12:50:31.550,San Francisco,,
+,
3,BEAR,2011-11-21 08:49:16.000,2011-11-21 08:49:19.987,San Francisco,,
+,
4,HAWK,2011-11-21 10:36:26.000,2011-11-21 10:45:11.823,Los Angeles,,,
5,HAWK,2011-11-21 10:36:26.000,2011-11-21 12:41:17.763,Los Angeles,,,
6,HAWK,2011-11-21 10:41:12.000,2011-11-21 10:55:08.393,San Francisco,,
+,
7,HAWK,2011-11-21 10:41:12.000,2011-11-21 15:46:24.707,San Francisco,,
+,
8,HAWK,2011-11-21 10:41:12.000,2011-11-22 11:09:13.907,San Francisco,,
+,

*** output
BEAR,San Francisco,4.02 hours
HAWK,San Francisco,24.47 hours
HAWK,Los Angeles,2.08 hours
[download]