But it's a requirement...

The one-liner above does it. Here's an expanded version of my one-liner that I hope makes it easier to understand.

#!/usr/bin/perl

# http://perlmonks.org/?node_id=1128822

use strict;
use warnings;

$_ = <<END; # input
1 H 35
1 C 22
2 H 20
2 C 30

A  1  HB2 MET 1
A  2  CA  MET 1
A  3  HA  ASP 2
A  4  CA  ASP 2
END

=output wanted
1  MET  HB2 35
1  MET  CA  22
2  ASP  HA  20
2  ASP  CA  30
=cut

#print "$1  $5  $4 $3\n" while /^(\S+)\s+(\w)\s+(\S+)(?=.*\n\n.*^\S+\s
++\S+\s+(\2..)\s+(\S+)\s+\1\b)/gms;

# expanded for clarity

print "$1  $5  $4 $3\n" while
  /       # match
  ^       # starting at the start of a line
  (\S+)   # capture first field
  \s+     # skip whitespace
  (\w)    # capture letter in column 2
  \s+     # skip whitespace
  (\S+)   # capture third field
  (?=     # zerowidth positive lookahead
  .*      # skipping to
  \n\n    # the empty line separating first and second file
          # this guarantees the patterns above this are in the first f
+ile
          # and the patterns below are in the second file
  .*      # skipping to
  ^       # start of a line in second file
  \S+     # skip first field (not needed)
  \s+     # skip whitespace
  \S+     # skip second field (not needed)
  \s+     # skip whitespace
  (\2..)  # capture third field if it starts with previously captured 
+letter (three wide)
  \s+     # skip whitespace
  (\S+)   # capture fourth field
  \s+     # skip whitespace
  \1      # make sure fifth field matches first field of first file.
  \b      # insure complete match
  )       # end of zerowidth lookahead
  /gmsx;  # global, match any start of line, . matches \n, expanded

__END__
[download]