Also have a look at Text::Table.

As for capturing the data, I think just a HoH will suffice. Your matrix presentation suggests that for each query and subject only one percent value is possible. In pseudo code:

my $matrix;
my %seen_queries;
for each row in the excel file:
  my ($query, $subject, $percent) = values from the excel row
  $matrix->{$subject}->{$query} = $percent;
  $seen_queries{$query} = 1;
end for
[download]

And to display the matrix:

use Text::Table;

my @cols = keys %seen_queries;
my $tb = Text::Table->new(@cols);
for my $subject (keys %$matrix) {
  $tb->load( [ @{$matrix->{$subject}}{@cols} ] );
}
print $tb;
[download]

This code demonstrates:

• the use of a hash to keep track of seen identifiers (%seen_queries)
• using a HoH to store a 2-D matrix of data
• hash slices to conveniently extract multiple values from a hash at once

Also, when writing this code I had to decide if I wanted to store the data as $matrix->{$subject}->{$query} or $matrix->{$query}->{$subject}. The latter organization would still work but would have made the table display code slightly more complex. Here's how it would have worked out:

use Text::Table;

my @cols = keys %$matrix;
my $tb = Text::Table->new(@cols);
for my $subject (keys %seen_subjects) {
  my @vals;
  for my $query {@cols) {
    push(@vals, $matrix->{$query}->{$subject});
  }
  $tb->load( [ @vals ] );
}
[download]

The hash %seen_subjects would be constructed in the reading phase just like %seen_queries but with $subject as the hash parameter.