I think you meant -ni.bak, not -pi.bak, there. As it stands you'll print out every line in the file once, and the interesting ones twice, which is suboptimal. :-)

As an alternative solution, if you know the number of lines in your header ahead of time, you can skip that number of lines using the flip-flop operator (..), thus:

#!/usr/bin/perl -w

# I'll assume for the sake of variety that we want to process the info
+rmation 
# in the file, not just delete the fluff
my $header_lines = 3;
while (<>) {
    next if 1 .. $header_lines;
    chomp;
    my @data = split;  
    &munge(@data);   
}
[download]

Note that I used split to divide the line up, but if the data fields are fixed-width and could contain embedded spaces, you'd be better off using unpack to split the line up (if they're variable-width and could contain embedded spaces, I recommend a different data format).