You could read in all of the data into an array of hash(references). A new hash(ref) would be added every time you encounter a UNIQUE-ID.

use strict;
use warnings;
use Data::Dumper;
my @parsed;
while(<DATA>){
    next unless /^(.*) - (.*)$/;
    push @parsed, { } if $1 eq "UNIQUE-ID";
    $parsed[-1]{ $1 } = $2;
}
print Dumper \@parsed;
__DATA__
#
UNIQUE-ID - GJDZ-5046
TYPES - BC-4
TYPES - Unclassified-Genes
COMMON-NAME - STM14_5042
ACCESSION-1 - STM14_5042
CENTISOME-POSITION - 90.96536    
COMPONENT-OF - CHROMOSOME-1-100
COMPONENT-OF - TUJDZ-2494
COMPONENT-OF - CHROMOSOME-1
LEFT-END-POSITION - 4430254
PRODUCT - GJDZ-5046-MONOMER
RIGHT-END-POSITION - 4430427
TRANSCRIPTION-DIRECTION - -
//
UNIQUE-ID - GJDZ-1101
TYPES - BC-4
TYPES - Unclassified-Genes
COMMON-NAME - focA
ACCESSION-1 - STM14_1100
CENTISOME-POSITION - 20.85712    
COMPONENT-OF - CHROMOSOME-1-23
COMPONENT-OF - TUJDZ-587
COMPONENT-OF - CHROMOSOME-1
LEFT-END-POSITION - 1015797
PRODUCT - GJDZ-1101-MONOMER
RIGHT-END-POSITION - 1016774
TRANSCRIPTION-DIRECTION - -
//
[download]

Duplicate entries would be overwritten though.