Here's an approach for the 'simple' input you give as an example. A larger chunk (but still a reasonable amount!) of more realistic input might yield a better solution. I see some other, similar postings from you – is there another thread on this with more data?

Notes:

• I use \x23 instead of a '#' character in the  [^-,^(\x23] character set below because of a peculiarity of my little command-line processor. You should just use '#' instead. (BTW: I'm not sure what all the '^' (carat) characters were doing in this set as originally posted, so I left one in there just for good luck!)
• The input text I use does not have accented characters. I can't display these easily on my console and so cannot test them.
• Because I do not use accented characters in my test input text, the regexes are untested with such characters. (Update: See Update 1 below.)
• As to the input text: Note that word 'Arranged' in the third record (i.e., line) has no preceding comma: 'Welsh Song Arranged'. Is this an example of real input, or a posting tyop? In any event, the code as it stands handles this variation.
• I concentrate on extracting what I take to be the critical fields from each record: the title of the piece and its source. You can stitch them together how you want, with commas, whitespace, whatever.
• Sorry for any wrap-around in the code listing.

>perl -wMstrict -le
"my @input = (
   'Tirana Alla Spagnola (Rossinizzatta), For Soprano & Piano (Peches 
+De Vieillesse, Book 1), Qr Iv/30',
   'A La Chapelle Sixtine, Transcribed, For Orchestra, S 360 (Lw G26)'
+,
   'All Through The Night, Traditional Welsh Song Arranged, For Mixed 
+Voices',
   );
 ;;
 my $not_source = qr{ [^-,^(\x23] }xms;
 my $kruft  = qr{ \s* ,? \s* }xms;
 my $ar_tr  = qr{ Arranged | Transcribed }xms;
 my $at_for = qr{
   $kruft $ar_tr? $kruft For $not_source+ $kruft
   }xms;
 my $rx_title = qr{ (?! $at_for) . }xms;
 ;;
 for (@input) {
   print qq{[[$_]]};
   my ($title, $source) =
     m{ \A \s* ($rx_title+) $at_for (.*?) \s* \z }xms
     ;
   print qq{:$title:  :$source:};
   }
"
[[Tirana Alla Spagnola (Rossinizzatta), For Soprano & Piano (Peches De
+ Vieillesse, Book 1), Qr Iv/30]]
:Tirana Alla Spagnola (Rossinizzatta):  :(Peches De Vieillesse, Book 1
+), Qr Iv/30:
[[A La Chapelle Sixtine, Transcribed, For Orchestra, S 360 (Lw G26)]]
:A La Chapelle Sixtine:  :S 360 (Lw G26):
[[All Through The Night, Traditional Welsh Song Arranged, For Mixed Vo
+ices]]
:All Through The Night, Traditional Welsh Song:  ::
[download]

Updates:

1. I have since tried this code as a regular source file with accented characters, and it seems to work.
2. Actually,  m{ \A \s* (.*?) $at_for (.*?) \s* \z }xms works just as well (for the limited test set), is probably a bit faster.

Okay:

#!/bin/perl

my $file='regex.txt';
 open my $fh, "<", $file
    or die "could not open $file: $!";
while (<$fh>) {
  chomp;
  if (  /(.*?)\s?(Transcribed|Arranged)?,\s+For\s+([^(,]+),?(.*)/) {
     print "$1$4\n";
     print STDERR "# Instrument: $3 T/A:$2\n";
  }
}
[download]

Using your input, ./program.pl 2>/dev/null output is:

Tirana Alla Spagnola (Rossinizzatta)(Péchés De Vieillesse, Book 1), Qr
+ Iv/30
À La Chapelle Sixtine, S 360 (Lw G26)
All Through The Night, Traditional Welsh Song
[download]