If you find yourself pulling out a rubber stamp then it is time to either use a subroutine or a data structure. In this case the key parts of the solution are to concatenate the ranges used for partitioning into an array and put the result of the partitioning into an array.

#!/usr/bin/perl
use strict;
use warnings;

my %netHash;
my @capVals;

my @limits = (1.0, 1.4, 1.9, 2.4, 3.0, 3.5, 4.0);

while(<DATA>) {
    chomp;
    my($nx,$ny) = split;
    $netHash{$nx} = $ny if ($ny > 1);
}

foreach my $key (keys %netHash) {
     my $value = $netHash{$key};
     my $index = BinSearch ($value, sub {$_[0] <=> $_[1]}, \@limits);
     
     next if $index < 0;
     $capVals[$index]{$key} = $value;
}

printHash ($_) for @capVals;

sub printHash {
  my $href = shift;
  
  return if ! defined $href;
  my $RandKey = (keys %{$href}) [int(rand(scalar (keys %{$href})))];
  print "$RandKey $href->{$RandKey}\n";
}


sub BinSearch {
    my ($target, $cmp, $array) = @_;
    my $posmin = 0;
    my $posmax = $#$array;
   
    return undef if ! @$array;
    return -0.5 if $cmp->($array->[0], $target) > 0;
    return $#$array + 0.5 if $cmp->($array->[-1], $target) < 0;
   
    while (1) {
        my $mid = int (($posmin + $posmax) / 2);
        my $result = $cmp->($array->[$mid], $target);
       
        if ($result < 0) {
            $posmin = $posmax, next if $mid == $posmin && $posmax != $
+posmin;
            return $mid + 0.5 if $mid == $posmin;
            $posmin = $mid;
        } elsif ($result > 0) {
            $posmax = $posmin, next if $mid == $posmax && $posmax != $
+posmin;
            return $mid - 0.5 if $mid == $posmax;
            $posmax = $mid;
        } else {
            return $mid;
        }
    }
}


__DATA__
      k1 0.2
      k2 0.9
      k3 1
      k4 1.2
      k5 1.3
      k6 2
      k7 2.2
      k8 2.7
      k9 2.76
      k10 3.001
      k11 3.12
      k12 3.2
      k14 3.22
      k15 3.3
      k16 3.34
      k17 3.4
      k18 3.62
      k19 3.8
      k20 3.89
[download]

Prints:

k4 1.2
k7 2.2
k9 2.76
k11 3.12
k19 3.8
[download]

This code assumes that there are not supposed to be gaps between the partitioned ranges. Upper and lower limit pairs would need to be stored and the search modified if the ranges truely are gappy.