How about using variables to turn confusion to clarity? Better yet, go to the library and borrow Perl Best practices to see why '\A' is better than \$'. spreading out your regex and adding comments, helps, too.

my $ham = "spam\tspam\tspam\tyam\tclam";
my $word    = qr{[^\t]+};
my $sep     = qr{\t};
my $capture = qr{($word)};
my (@jam ) = ($ham =~ m{\A                 # enforce beginning of stri
+ng
                         $word $sep        # skip first word and separ
+ator
                         $word $sep        # and the second
                         $capture $sep     # capture next two words
                         $capture $sep     # skip the separators
                         $word             # skip a word
                        \z                 # and then it's the end of 
+the string
                       }xms);
print join("\n", '**', @jam, '**', '');
[download]

Using the debugger helps, too ... Along the way I noticed your string has a double tab '\t\t', but you only ever accept single tabs; You specify end of string '$', when there's still another word to go.

But you're doing too much work. You could split() on '\t' and select only the components you want. If it's the 3rd % 4th ...

my @jam = ( spit "\t", $ham )[2,3];
[download]

If you do need to use a regex, do you need to check whether there is a word after your capture? Do you need to enforce there is nothing after that last word? Simplify your regex for greater happiness.