You can avoid the need to sort or even do any comparisons. Just loop through the data and build a string using one character to represent your ranges and another the 'freespace'. Then just count the non-freespace chars and your done.

#! perl -slw
use strict;

my @data = map{ [ split "','", substr( $_, 1, -2 ) ] } <DATA>;

my $mask = ' ' x 10_000; # Make a blank string that's 'big enough'
my $max = 0; # we'll trim it back later

for( @data ) {
      #calc the length of the range
    my $len = $_->[2] - $_->[1] +1;

      # overwrite the range with 'xx's
    substr( $mask, $_->[1], $len ) = 'x' x $len;

      # Remember the highest offset
    $max = $_->[2] if $_->[2] > $max; 
}

$mask = substr( $mask, 0, $max ); # Trim to length

my $coverage = $mask =~ tr[x][x]; # Count the 'x's

printf "cover = %.1f%% \n", $coverage  / length($mask) *100;

__DATA__
'NM_176827','618','710'
'NM_176827','621','710'
'NM_176827','622','692'
'NM_176827','629','710'
[download]

Output

P:\test>279587
cover = 13.0%
[download]