My approach is to filter records while reading data.

use warnings;
use strict;

my $format = "%-11s %-8s\n";

open my $data, '<', 'data.txt'
    or die "Could not open 'data.txt': $!\n";

my %dataset;

RECORD:
while ( my $line = <$data> ) {
    chomp $line;
    my ( $coord, $dist ) = split ',', $line;
    next RECORD
       if exists $dataset{$coord} and $dataset{$coord} < $dist;
    $dataset{$coord} = $dist;
}

printf $format, 'coordinate', 'distance';
printf $format, $_,  $dataset{$_}
    for sort { $a <=>  $b } keys %dataset;
[download]

• I turn on warnings and strict as soon as my program grows beyond a few lines, because I don't need to spend 2 hours investigating something strange which turns out to be a typo.
• I'm not sure how specific your format needs to be, so I assumed it was hard-coded column widths and used printf with a format string , rather than print or say.
• Read in lines from a comma-separated file, discard the newline and split on comma into temporary variables.
• If the distance is not the shortest on record for this coordinate, jump to the next input record. Although you can use next or last or redo without a label, I like to always specify what I'm next-ing or last-ing or redo-ing, especially if the loop is long or if it's a nested loop. Save the data if we haven't bailed.
• Extract the keys from the hash, sort them numerically, and print out the keys and values.
• I never actually close input files ... this case is too short and fast to care, and in more complicated instances I would have a readfile() routine, which would automatically close the file when it went out of scope.
• You can get more information on next, last, sort and other stuff by typing perldoc -f sort or whatever the command name may be, into an xterm window, or you can use the perldoc web site