AnonyMonk Update: and JediWizard have accurate answers has an accurate answer </update> for the question you appear to be asking but perhaps the question you intended is about why $& has values.

If so, note that $& is a special regex variable that -- quoting from Chapter 7 in "Mastering Regular Expressions" (page 299 in my 2nd Ed. paperback) -- "A successful match or substitution sets a variety of global, read-only variables that are always automatically, dynamically scoped. These values never change if a match attempt is unsuccessful, and are alwaysset when a match is successful." (emphasis in the original; but note especially the first clause of the last sentence)

Update#2 (See AnomalousMonk's below): Your line 8, if(~/^<dsk1/) matches on "<dsk1" and therefore sets $&. Your line 8 merely tests whether bitwise negation of the string is possible. Since there is no match the prior match is retained in $& (I think). </Update#2> Execution (modified code below) produces this:

perl 851032-orig.pl
$& at line 9: <dsk1
 **********
1

$& at line 14: <dsk1 line1 id="123" lo="1" to="abc" rb="This is a long
+ rb."
 ------------------
$& at line 9: <dsk1
 **********
2

$& at line 14: <dsk1 line2 id="456" lo="2" to="def" rb=Short rb"
 ------------------
$& at line 9: <dsk1
 **********
3

$& at line 14: <dsk1 line3 id="789" lo="3" to="ghi" rb="Medium long rb
+"
 ------------------
$& at line 9: <dsk1
 **********
Use of uninitialized value in concatenation (.) or string at 851032-or
+ig.pl line 13, <DATA> line 4.


$& at line 14: <dsk1
 ------------------
$& at line 9: <dsk1
 **********
5

$& at line 14: <dsk2 line5 id="555" lo="5" to="jkl" rb="should not mat
+ch"
 ------------------
$& at line 9: <dsk1
 **********
6

$& at line 14: <dsk1 Line6 id="987" lo="6" to="mno" rb="This should be
+ a match"
 ------------------
$& at line 9: <dsk1
 **********
7

$& at line 14: <dsk1 line7 id="FFF" lo="7" to="pqr" rb="This is a very
+, very, very long are-bee."
 ------------------
[download]

using this, slightly modified code:

#!/usr/bin/perl
use strict;
use warnings;
# 851032-orig (but using data)

# open FH,"data" or die "can't open the file";

while(<DATA>) {
    if(~/^<dsk1/) {
        print "\$& at line 9: $& \n **********\n";    # added
        my $line=$_;                                  # added
        $line=~/.*\sid=\"(.*)\"\slo=\"(.*)\"\sto=\"(.*)\"\srb=.*$/;
        print "$2\n\n";
        print "\$& at line 14: $&\n ------------------\n";
    }
}
__DATA__
<dsk1 line1 id="123" lo="1" to="abc" rb="This is a long rb."
<dsk1 line2 id="456" lo="2" to="def" rb=Short rb"
<dsk1 line3 id="789" lo="3" to="ghi" rb="Medium long rb"
<dsk2 line4 
<dsk2 line5 id="555" lo="5" to="jkl" rb="should not match"
<dsk1 Line6 id="987" lo="6" to="mno" rb="This should be a match"
<dsk1 line7 id="FFF" lo="7" to="pqr" rb="This is a very, very, very lo
+ng are-bee."
[download]

Note also that you are using conventional regex notation at line 12 but not in line 9. Update2: I'm unclear why line 9 passes a syntax check... but that just means I have more fun hunting up the answer on that. (As noted, Anomalous Monk answers below.)

Hence, I'm posting code and output with the match syntax (rather than bitwise negation):

#!/usr/bin/perl
use strict;
use warnings;
# 851032

#open FH,"data" or die "can't open the file";

while(<DATA>) {
    my $line=$_;
    if( $line =~ /^<dsk1/ ) {
        print "\$line: $line";
        if ($line =~ /.*\sid=\"(.*)\"\slo=\"(.*)\"\sto=\"(.*)\"\srb=.*
+$/) {
            print "\$1: $1, \$2: $2\, \$3: $3 \n";
            print "-----------------\n";
        } else {
            print "Some of this will be uninitialized: ";
            print "\$1: $1, \$2: $2\, \$3: $3 \n";
            # print "\$&: $& \n\n---\n";
        }
    } else {
        print "\$line did NOT start with '^dsk1': $line \n ===========
+======\n\n";
    }
}

=head
Output:  (see also 851032-orig.pl)

perl 851032.pl
$line: <dsk1 line1 id="123" lo="1" to="abc" rb="This is a long rb."
$1: 123, $2: 1, $3: abc
-----------------
$line: <dsk1 line2 id="456" lo="2" to="def" rb=Short rb"
$1: 456, $2: 2, $3: def
-----------------
$line: <dsk1 line3 id="789" lo="3" to="ghi" rb="Medium long rb"
$1: 789, $2: 3, $3: ghi
-----------------
$line did NOT start with '^dsk1': <dsk2 line4

 =================

$line did NOT start with '^dsk1': <dsk2 line5 id="555" lo="5" to="jkl"
+ rb="should not match"

 =================

$line: <dsk1 Line6 id="987" lo="6" to="mno" rb="This should be a match
+"
$1: 987, $2: 6, $3: mno
-----------------
$line: <dsk1 line7 id="FFF" lo="7" to="pqr" rb="This is a very, very, 
+very long are-bee."
$1: FFF, $2: 7, $3: pqr
-----------------

=cut

__DATA__
<dsk1 line1 id="123" lo="1" to="abc" rb="This is a long rb."
<dsk1 line2 id="456" lo="2" to="def" rb=Short rb"
<dsk1 line3 id="789" lo="3" to="ghi" rb="Medium long rb"
<dsk2 line4 
<dsk2 line5 id="555" lo="5" to="jkl" rb="should not match"
<dsk1 Line6 id="987" lo="6" to="mno" rb="This should be a match"
<dsk1 line7 id="FFF" lo="7" to="pqr" rb="This is a very, very, very lo
+ng are-bee."
[download]

</update2a>

A few other suggestions:

• You can be far more precise and confident of your results if your id, lo, and to are consistent by replacing dot-star (aka death-star for the trouble it can bring you) with an appropriate pattern.
• Testing your regex results is at least (IMO) as important as testing a file_open, which you've done (albeit in 2-argument form, vice the preferred 3-argument open)
  for example,  if ($1 ) { do something with it; } else { report the failure; }