The code seems overly complex to me.
There are some steps:
a) parse the input lines and assign values to some variables.
b) Use those variables to create the data structure.

The below could be a way to parse the data.
A HoHoH is a pretty complex data structure. What are you trying to do? There might be some better alternatives, but I don't know enough to comment on that. Can you explain more about your objective?

#!usr/bin/perl -w
use strict;

my %site_info=();

while (<DATA>)
{
  next if /^\s*MULTI/;  #skip MULTI lines
  next if /^\s*$/;      #skip blank lines
  
  # default split is /\s+/ which includes
  # all of the whitespace charaters [ \n\r\f\t]
  
  # From the data shown (no spaces within the
  # tokens), why not just split on simple whitespace?
  
  my ($chrStr, $strand, $value) = (split)[0,1,3];  # ditch the dash
  
  # use simple regex to get the "chr number"
  my ($chrNum) = $chrStr =~ /(\d+)/;  #just get the digits..
  
  print "$chrNum $strand $value\n";
}
    
=prints
16 28797601 0.777877
32 887755 0.11
=cut

    
__DATA__
chr16 28797601 - 0.777877
MULTI some stuff
chr32 887755 - 0.11
[download]