OK I dragged out an old book.

Here it goes under the heading of "Trans or Cis effects".

"Both A and B genes can be inherited as one unit on the same chromosome,
known as Cis-AB. In this rare exception , an AB parent may have an O child.,
normally considered as exclusion of maternity or paternity.
This variant can be recognized by both the weakness of the B antigen and a
weak anti-B present in the serum."

1: Vox Sang. 2004 Jul;87(1):41-3. Related Articles, Books, LinkOut
The serological and genetic basis of the cis-AB blood group in Korea.

Cho D, Kim SH, Jeon MJ, Choi KL, Kee SJ, Shin MG, Shin JH, Suh SP, Yazer MH,
Ryang DW.

Department of Laboratory Medicine, Chonnam National University Hospital,
Gwangju, South Korea.

BACKGROUND AND OBJECTIVES: The cis-AB blood group is rare, although
relatively common amongst Koreans. The serological characteristics and
genetic basis of Korean cis-AB blood donors were investigated. MATERIALS AND
METHODS: Polymerase chain reaction-restriction fragment length polymorphism
(PCR-RFLP), designed to detect the cis-AB01 allele, was performed on 194 AB
samples which demonstrated weak or unusual expression of either or both of
the A or B antigens. RESULTS AND CONCLUSIONS: Sixty cis-AB01 donors were
identified. cis-AB01/O01 or O02 were the most common genotypes (36/60)
detected only in A(2)B(3) donors, and cis-AB01/B101 (nine of 60) was the
least common genotype identified only in A(2)B donors. Surprisingly
cis-AB01/A102 (15/60) was identified in a variety of phenotypes (A(1)B(3),
A(1)B(x) or el, A(int)B(3)).

PMID: 15260821 [PubMed - indexed for MEDLINE]

This under the heading of "mutations".

"Theoretically, the presence of a genetic marker in a child not found in
both parents can be explained on the basis of mutation. However, the
mutation rate in man has been estimated as close to one in a million, and no
convincing example of a mutation involved in paternity testing has ever been
found."

With regards to Bombay et al

"If the regulator gene is absent, not functioning properly, or being
suppressed, there is no precursor substance being formed for the structural
gene to act upon, with the end result of an incomplete or non-detectable
genetic marker. In this case, the normal structural gene can be passed on to
the child. If the child then inherits a normal regulator gene from the other
parent, a normal genetic marker will be produced. The Bombay phenotype, Rh
null, and McCloud phenotype can all be explained by this hypothesis.

"In the ABO system, the H substance has been shown to be the precursor of A
and B substances. Thus in Bombay phenotype, the regulator H gene is absent,
while the A and B structural gene capable of transforming precursor  H
substance to A or B substance is normal; in a group O person, both the A and
B genes are absent"

"It is possible for two seemingly group O individuals to have a group A or B
child, just as it is possible for two Rh null individuals to have a child
with normal Rh antigens."

Bottom line in the past 20 blood group antigens were used in paternity
testing and not just ABO.

Now the exclusion rate for DNA testing is so large that there is no
comparison.

Unless I'm misunderstanding  how the Bombay phenotype works (which is
certainly possible), that's not correct. One parent could be AB and Hh
and the other could be OO and Hh. The child can have a genotype of AO
(or BO) and hh and will appear phenotypically to be O because the hh
will prevent expression of the A or B proteins.