A "greedy" match will indeed get greedy, but it will always allow the last part of the pattern to match if that is possible. Adding a .* at the beginning of the pattern allows that .* to "gobble up" the first START while allowing for the last START to match up with some characters followed by END.

#!/usr/bin/perl -w
use strict;

my $text ="some text START text I don't want START only text I want EN
+D";

my $wanted  = ($text =~ /.*START (.*) END$/)[0];
my $wanted2 = ($text =~ /.*(START .* END)$/)[0];

print "wanted=\"$wanted\"\n";
print "wanted2=\"$wanted2\"\n";
__END__
prints:
wanted="only text I want"
wanted2="START only text I want END"
[download]

Update: this: my $wanted  = ($text =~ /.*START (.*) END$/)[0]; may look a bit strange, but this is how to assign $1 to $wanted without having to use $1 as an intermediate variable. The text match is in a list context and I just slice to get the contents of the first matching paren. $2 can be done in the same way...

my ($x,$y) = ($text =~ /.*(START (.*) END)$/)[0,1];
print "x=$x y=$y\n";
#prints: x=START only text I want END y=only text I want
[download]

I like this syntax as it "gets to the point" without $1,$2,$3, etc.