You're basically going to have to quantify the different possibilities and allow for them individually. I was able to get the zip codes accurately from your sample data:

 unless ( ($zip) = ($field5 =~ /(\d{5}-\d{4})/)) {
                unless ( ($zip) = ($field5 =~ /(\d{5})/)) {
                        unless ( ($zip) = ($field4 =~ /(\d{5}-\d{4})/)
+) {
                                ($zip) = ($field4 =~ /(\d{5})/);      
+   
                        }                                       
                }        
        }
[download]

but that's already pretty nasty...