Here's another approach. The module is extremely simple: no exportation, no OO. Invocation must be by fully-qualified function names, but that's part of the simplicity.

The idea is that the regexes used for record field validation and extraction can be adapted very precisely to the data. (These regexes are currently rather general and naive; names and titles and such can be very tricky.) No effort is made to check for play index number/name duplication. The test script should be expanded to cover many edge cases. Note that the "Polonius" output field of play 46005 has been "fixed": The trailing space present in the OP has been removed; if this whitespace needs to remain, you need to specify under what conditions trailing whitespace is preserved.

c:\@Work\Perl\monks\jcg3525>perl ParsePlaybill.t
ok 1 - use ParsePlaybill;
ok 2 - ^46004 % Tamerlane.| Tamerlane - Sheridan; Bajazet - Barry; Mon
+eses - A Gentleman; Arpasia - Mrs. Furnival; Selima - Mrs. Elmy; +
#
# Polonius "fixed" in this test case
ok 3 - ^46005 % Hamlet.| Hamlet - Sheridan; Polonius - J. Morris; Laer
+tes- Lacy; Ophelia- Mrs. Storer; Queen - Mrs. Furnival; +
#
# one-character play
ok 4 - ^1 %   Mark Twain Tonight  .| Mark Twain  -Hal Holbrook;+
#
ok 5 - no warnings
1..5
[download]

Update: As it stands, this code has some (small, I hope) problems:

1. After looking at some of the posts of holli and Marshall, I realized that character names with embedded hyphens would cause a problem (conflict with the character-actor field separator hyphen). There's a simple fix for this, but it's fragile.
2. Names of a play, character or actor (player) can include  - ' . characters after the first character, but cannot end in one of these characters due to the  \b assertion in the regexes for these record fields. Again, I think there's a simple fix, but it may be fragile. (These fields probably could not end in a hyphen in any case.)

And further to this point: Falsehoods Programmers Believe About Names.

---

Give a man a fish:  <%-{-{-{-<

This is a completely valid point.
With names there is of course always the additional complications of titles, XXX,MD or Jr. Sr., III etc. Sometimes people have four or just two names instead of three. Wild cases like a single name "Madonna" also happen. A friend of mine didn't get fully divorced, but did drop her last name (which she got from her husband). A weird thing short of going back to her maiden name.

In my code at Re: Help formatting text to delimited text in file, my line my ($name1, $name2) = split /\s*-\s*/,$sub_name; was my ($name1, $name2) = split /\s+-\s+/,$sub_name; until I ran it and saw: Ophelia- Mrs. Storer instead of the expected Ophelia - Mrs. Storer.

I am not sure if the missing space before the "-" is a typo or not? I changed the split regex to allow optional spaces before and after to work with the OP's posted data in a quick decision. Usually a hyphenated name will be printed without spaces either before or after each surname. Mileage varies.

With just 2 example lines, we can't solve every potential case. There is always some iteration involved when working ad-hoc without a complete spec. It could be that requiring a space after the "-" is enough to differentiate between "Smith-Jones"? Not sure.

I think my suggestion to count hyphens on each line and identify outliers is a good one. Modify code accordingly.
I still suspect that: "Ophelia- Mrs. Storer" is a typo.
Update: Oh, in this type of printout, I sincerely doubt that that there are any "escape" characters like "\" to guide the process. Could be, but doubtful. I am working on a project right now where I have to parse several types of printouts designed for humans. Perl is an excellent language for this! Regex is an operator instead of an object and I can quickly iterate and fine tune the parsing functions.

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

open my $IN,  '<', 'input.txt'  or die "Cannot open 'input.txt' becaus
+e: $!";
open my $OUT, '>', 'output.txt' or die "Cannot open 'output.txt' becau
+se: $!";

# Input record separator
$/ = '+';

while ( <$IN> ) {
    # skip empty records
    next unless /\S/;
    # remove input record separator
    chomp;
    my ( $id, $name ) = ( $1, $2 ) if s/ \s* \^ \s* ( \d+ ) \s* % \s* 
+( .+? ) \s* \.\| \s* //x;
    for my $data ( split /\s*;\s*/ ) {
        print $OUT join( ',', map qq/"$_"/, $id, $name, split /\s*-\s*
+/, $data ), "\n";
        }
    }
[download]

What did you code so far?

Hint: Looks like you need to apply multiple rounds of split , i.e. first on | then ; then -

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

while (my $line = <DATA>)
{ 
   if (my( $num, $root_name, $rest)=$line =~ /^\^(\d+)\s*%\s*(\w+)[. |
+]+(.+)\+$/)
   {    
       my @sub_names = split /;\s*/,$rest;
       
       foreach my $sub_name (@sub_names)
       {
           my ($name1, $name2) = split /\s*-\s*/,$sub_name;
           
           print "\"$num\",\"$root_name\",\"$name1\",\"$name2\"\n";  
       }
   }
}
=Prints
"46004","Tamerlane","Tamerlane","Sheridan"
"46004","Tamerlane","Bajazet","Barry"
"46004","Tamerlane","Moneses","A Gentleman"
"46004","Tamerlane","Arpasia","Mrs. Furnival"
"46004","Tamerlane","Selima","Mrs. Elmy"
"46005","Hamlet","Hamlet","Sheridan"
"46005","Hamlet","Polonius","J. Morris"
"46005","Hamlet","Laertes","Lacy"
"46005","Hamlet","Ophelia","Mrs. Storer"
"46005","Hamlet","Queen","Mrs. Furnival"
=cut


__DATA__
^46004 % Tamerlane.| Tamerlane - Sheridan; Bajazet - Barry; Moneses - 
+A Gentleman; Arpasia - Mrs. Furnival; Selima - Mrs. Elmy; +

^46005 % Hamlet.| Hamlet - Sheridan; Polonius - J. Morris; Laertes- La
+cy; Ophelia- Mrs. Storer; Queen - Mrs. Furnival; +
[download]

#!/usr/bin/env perl

use Data::Dumper;
use 5.14.0;

my @data=(
        q{^46004 % Tamerlane.| Tamerlane - Sheridan; Bajazet - Barry; 
+Moneses - A Gentleman; Arpasia - Mrs. Furnival; Selima - Mrs. Elmy; +
+}
        ,q{^46005 % Hamlet.| Hamlet - Sheridan; Polonius - J. Morris; 
+Laertes- Lacy; Ophelia- Mrs. Storer; Queen - Mrs. Furnival; +}
        );
local $"='","';#"
for (@data) {
    my %c_h;
    if (m{^\^(.+?) % (.+?)\.\|(?{$c_h{1}=$1; $c_h{2}=$2;}) ?(?:(.+?) *
+- *(.+?); ?(?{$c_h{3}=$3;$c_h{4}=$4; say qq{"@c_h{qw(1 2 3 4)}"}}))*}
+) {
        };
    };
__DATA__
[download]

"46004","Tamerlane","Tamerlane","Sheridan"
"46004","Tamerlane","Bajazet","Barry"
"46004","Tamerlane","Moneses","A Gentleman"
"46004","Tamerlane","Arpasia","Mrs. Furnival"
"46004","Tamerlane","Selima","Mrs. Elmy"
"46005","Hamlet","Hamlet","Sheridan"
"46005","Hamlet","Polonius","J. Morris"
"46005","Hamlet","Laertes","Lacy"
"46005","Hamlet","Ophelia","Mrs. Storer"
"46005","Hamlet","Queen","Mrs. Furnival"
[download]

#!/usr/bin/env perl

use Data::Dumper;
use DBI;
use 5.14.0;
use if ($ENV{DBG} || $ENV{DEBUG}),'Devel::Include','keep',$ENV{MASK} ?
+ $ENV{MASK} : '#=#';

my @data=(
        q{^46004 % Tamerlane.| Tamerlane - Sheridan; Bajazet - Barry; 
+Moneses - A Gentleman; Arpasia - Mrs. Furnival; Selima - Mrs. Elmy; +
+}
        ,q{^46005 % Hamlet.| Hamlet - Sheridan; Polonius - J. Morris; 
+Laertes- Lacy; Ophelia- Mrs. Storer; Queen - Mrs. Furnival; +}
        );
my $dbh=DBI->connect('dbi:SQLite:dbname=test.sqlite','','',{ PrintErro
+r=>1, RaiseError=>1 });
$dbh->do(<<"__CREATE__");
    CREATE TABLE titles (
        number integer,
        name text);
__CREATE__
$dbh->do(<<"__CREATE__");
    CREATE TABLE roles (
        number integer,
        role text,
        actor text);
__CREATE__
my %sth;
$sth{title}=$dbh->prepare(<<"__TITLE__");
    INSERT INTO titles (number,name) VALUES (?,?);
__TITLE__
$sth{role}=$dbh->prepare(<<"__ROLE__");
    INSERT INTO roles (number,role,actor) VALUES (?,?,?);
__ROLE__
local $"='","';#"
for (@data) {
    if (m{^\^(.+?) % (.+?)\.\|(?{$sth{title}->execute($1,$2)}) ?(?:(.+
+?) *- *(.+?); ?(?{$sth{role}->execute($1,$3,$4)}))*}) {
        };
    };
$dbh->disconnect;
[download]