That's what I don't get either. He says in the one sample found that police "should" re-test the "matches" using the SGMplus technique adding the samples would "likely" be eliminated after this was done. I don't have an issue with that statment, the SGMplus test used at the time I believe used 10 profile markers, as opposed to the old SGM test which only used 6. So statistically speaking, once they got data for the other 4 markers, it was more "likely" than not that there would no longer be a match. That's a statement of mathematical fact but it's not to say they wouldn't still match either, hence the recommendation to investigate further making absoute sense. Otherwise, why bother checking any of the profiles on the old SGM system at all if you are just then going to discount any matches?
But then in the second sample they found, he doesn't recommend carrying out an SGMplus test. He says that "as that sample was processed with the old SGM system I consider the match to be of negligible significance". But what is the difference here? Both samples went through the same elimination process and both still showed "matches" with profiles on the database based on the 6 markers common to both tests. So why recommend carrying out re-tests for the one but then state "I consider" the second sample not to matter. Based on what? The only reason he gives is that it was from the old SGM system... but so was the other one.
If there's more to it than that, he should state the reasons in the report really. Maybe it's because the original sample wasn't linked to an actual individual, only another crime scene. So perhaps it wasn't possible to re-test it using the newer technique because the original sample had been destroyed whereas for the other one, I suppose they could approach the individual in question to get a DNA swab.