I've stored your data in a file called 782426.txt, because I couldn't get DATA to work with embedded Ctrl-Z characters.

Now, in the below solution, I read the data a line at a time, building up a record until I encounter the line which I know, a priori, to be the last one in each record. I could have tried to exploit the blank line which occurs between records, but in dealing with the header, I rather crudely blow away all blank lines.

use strict;
use warnings;

open F, '<', '782426.txt' or die;
$/="\r\n";
binmode F;

my $last_key;
my $record = {};
my @records;
while (<F>)
{
    s/^//; # kill that pesky thing.
    /^\s+NEW CAR INV prepared by / and scalar(<F>), next; # the header
    /^[\.\s]*$/ and next; # skip any blank lines
    /^\d+ records listed\./ and last; # end of report
    chomp;
    my($key,$val) = /^(.{19}) +(.*)/;
    if ($key=~/\S/)
    {
        $key =~ s/\.+$//; # kill trailing dots
        $record->{ $last_key=$key } = $val;
    }
    elsif(defined $last_key)
    {
        $record->{$last_key} .= $val;
    }
    if ( $key eq 'SALES CST' ) # last line of record
    {
        push @records, $record;
        $record = {};
    }
}

close F;

for my $record ( @records )
{
    print "$_='$record->{$_}'\n" for sort keys %$record;
    print "\n";
}
[download]

I didn't bother trying to deal with that line at the top of your data which contains a lone '1' character.