I don't know if your code is correct or not, but it's certainly more complex than it needs to be.

Here's how I'd built the frequency statistics:

use POSIX qw(floor);
sub build_histogram {
   my ($bucket_size, @items) = @_;
   my %result;
   for (@items) {
       my $bucket = $bucket_size * floor($_ / $bucket_size);
       $result{$bucket}++;
   }
   return %result;
}
[download]

This sorts everything from 0 to $bucket_size (exlusively) into the first bucket labeled 0 etc. If you want it to be sorted into a bucket labeled $bucket_size instead, use ceil() instead of floor().

Anyway, it's hard to tell if your code is correct or not. Try to use Data::Dumper to check if the intermediate data structures are correct, and perhaps even do it manually for a short excerpt of the log files and compare the results.