Heart attack time

Two weeks ago I was wondering when the deadline for our annual Quality Control tests was coming up, so I looked it up, and discovered - omg!- it was in four days.
Ummm. I hoped the results were ready for writeup, because 4 days is not enough time to start from the beginning.
Well, yes, everything was fine. Just not filed where I expected it, which was why I hadn’t been alerted when the tests were complete. A small bug to fix.
The result for colon cancer case 2, though, went down to the wire. It was a tricky one, and used a technique that we’re not always successful with. We ran the experiment five times, in duplicate, triplicate, or more, each time. In retrospect, we had it on the third try. At the end I was confident in our result and happy with the report - single exon deletions are the hardest to be sure of.
.
Today, our reports have not been graded yet, but the official results of what we were supposed to find have just come out. Breast cancer case 1, check; case 2, check; case 3, check. So far so good. Colon cancer case 1, check; case 2...
Case 2...
WTF??
c.994dupA? a point mutation?

.
Ah, no. (heart goes faster)
.

No, no, that’s not what we found. (heart goes faster and a bit wobbly)
How could we be so wrong? (please stay seated)
.

Heywaitaminute. Take another look at the instructions for the case. Yes, I thought so: ...
"...Please perform mutation analysis in MLH1 starting by a search for large genomic deletions and duplications in MLH1. For the purpose of this scheme participants are not required to carry out full sequence analysis of all the exons of the MLH1 gene."
That was pulled off the web site two weeks ago and is slightly but importantly different than the version I printed the day we received the samples, in which the second sentence was: "Sequencing analysis failed to detect a mutation."(my bold).

So naturally we did no sequencing, which will pick up small mutations with no problem at all.

Ergo:
The quality control people have made an error in posting the expected results.
At least I hope so.
.
Otherwise it’s quite complicated, but it could happen. All sorts of very strange things happen if you just keep looking. Look at a million cases, and you’ll find plenty that are one-in-a-million.

(here's where I get excessively technical)
See, our technique relies on using a short probe unique to each exon of the gene being analyzed. We hybridize the collection of probes to the DNA of the patient, and get a quantitative answer about how many copies of that exon the patient has. Two is normal. But if the duplication of a single base just happens to land in a sensitive part of our probe, it could interfere with the hybridization and give us a false result. It could tell us there’s only one copy of the exon, when in fact there are two but one of them is no longer recognized by the probe. These probes are part of a commercialized kit and are supposedly resistant to changes at a single base, but, well...

.
So I’ve sent a question off to the QC people, who have not yet responded. And I’m having the lab do the sequence of the exon in question, to see if this c.994dupA is there or not. If it is, we’ll have to lodge a complaint about being instructed to look for a particular kind of mutation, the techniques for which are normally insensitive to finding the mutation in question (not just our technique, but two others commonly used should also miss this).

.