And how is '' => '}' replacing the line ending? I know \z matches the end, but those empty 'quotes' puzzle me.

That's easy enough to explain. In essence, it's a bit of regex affectation. The idea was to stick a }-curly on the end of a string. That can be done well enough with a
    $string .= '}';
statement, but I thought it would be neat to it with the same global substitution that was handling all the other transformations. The empty pattern matches the empty string, and the empty string is present n+1 times in a string of n length.

c:\@Work\Perl\monks\tel2>perl -wMstrict -le
"my $s = '123';
 $s =~ s{ }{X}xmsg;
 print qq{A: '$s'};
 ;;
 $s = '';
 $s =~ s{ }{X}xmsg;
 print qq{B: '$s'};
 ;;
 $s = '123';
 $s =~ s{ \z }{X}xmsg;
 print qq{C: '$s'};
"
A: 'X1X2X3X'
B: 'X'
C: '123X'
[download]

And there's always an empty string at the  \z (absolute end) of any string, even the empty string! (And the empty string is a perfectly good hash key.)

As you have seen, this approach breaks down when you try to replace an empty string at different locations with different replacement strings: there can only be one ''-keyed value in a hash. To do what you want with a regex, you'd have to resort to some screaming hack like (untested)

my %globize = ('P' => '{P',  ':' => '}:{', '' => '}');
...
$globule =~ s{ (\A P | : | \z) }{$globize{$1}}xmsg;
[download]

(in the unlikely case your input string always starts with a fixed sequence like 'P'), but if you just want to add different stuff at the start and end of a string, it would be better IMHO to just do something like
    $globule = "{$globule}";
instead (update: and forget about all the  \z alternation and  '' => '}' stuff).