Perhaps the following data structure--a hash of hashes of arrays (HoHoA)--would be helpful (as suggested in your posting's topic):

use strict;
use warnings;
use Data::Dumper;

my %chrHash;

push @{ $chrHash{"chr1"}{3} }, 100;
push @{ $chrHash{"chr1"}{3} }, 300;

print 'Count for chr1 3: ', scalar @{ $chrHash{"chr1"}{3} }, "\n";
print qq/Locations for chr1 3: @{ $chrHash{"chr1"}{3} }\n\n/;

print Dumper \%chrHash;
[download]

Output:

Count for chr1 3: 2
Locations for chr1 3: 100 300

$VAR1 = {
          'chr1' => {
                      '3' => [
                               100,
                               300
                             ]
                    }
        };
[download]

The arrays would track both count and locations.

On another issue, consider:

$chrHash{$_} = { %countHash } for @chromosomes;
[download]

instead of using map in a void context, like:

map { $chrHash{$_} = { %countHash } } @chromosomes;
[download]

Also, consider using grep or List::Util's first instead of the smart-match operator (see Smart matching is experimental/depreciated in 5.18 - recommendations?):

use strict;
use warnings;
use List::Util qw/first/;

my @chromosomes = ( "chr1", "chr2", "chr3", "chr5" );
my @line = ( "chr1", "chr3", "chr5" );

if ( grep $line[2] eq $_, @chromosomes ) {
    print "Found $line[2] using grep!\n";
}

if ( first { $line[2] eq $_ } @chromosomes ) {
    print "Found $line[2] using first!\n";
}
[download]

Output:

Found chr5 using grep!
Found chr5 using first!
[download]

The advantage of first is that it terminates upon the first successful find, whereas grep will iterate through the entire list.