I have thought about this approach. I have already written the first part (the data is stored in the hash %details of match):

my @output_lines;
my $model_number=0;
foreach my $gene_name (keys %details_of_match) {
    
    $output_lines[$model_number]= qw{ } x length($original_seq_string)
+;
    
    for (my $i=0; $i< scalar( @{$details_of_match{$gene_name}} ); $i++
+ ) {
        
        substr       $output_lines[$model_number]         # the new li
+ne
                , $details_of_match{$gene_name}->[$i][0]        # the 
+position
                , 21     # the length (constant)
                , $details_of_match{$gene_name}->[$i][2]     #the patt
+ern
                ;
    }    
    $model_number++;
}
[download]

At this point, I have all the lines, each one with just one substring in the right position, stored in the array @output_lines. But then, I get stuck with the second part, because I don't imagine how to collapse all the lines into the least number of possible lines.