Well no, I'm not sure that I got it. What is clear is that you did not satisfy GrandFather's request in the first reply, as I hoped you would.

So let me make another guess at what you really want. How about this:

my @arr = ();
while (<PIR>)
{
    chomp;
    if( /^ENTRY/ ) { $entry = $_ }
    elsif ( /^(TITLE)\s+(\S.*)/ ) { $title = "$1\n\t $2" }
    elsif ( /^(ORGANISM)\s+(\S.*)/ ) { $org = "$1\n\t $2" }
    elsif ( /^ACCESSIONS/ ) { $acc = $_ }
    else {
        push @arr, $_;
    }
}

print "@arr\n";
[download]

Now, I would assume there should be more code than that, if you really need to do things with $acc, $entry, $org and $title. If you really just want to output an array with those long strings as the elements of the array, the code could be a lot simpler.

If there's a chance that one of those long strings might appear more than once in the data file, use those long strings as hash keys instead of array values:

# simplified version: ignore header stuff:

my %hash;
while(<PIR>)
{
    chomp;
    $hash{$_} = undef unless /^(?:ENTRY|TITLE|ORGANISM|ACCESSIONS)\s/;
}

print join " ", keys %hash, "\n";
[download]

Using a hash like that might be a good idea for other reasons: maybe you would want the header values to be associated with each long string. (Hint: some people refer to hashes as "associative arrays".) If so, assign the header strings as the hash value.