Ulyssus

I checked this and indeed, there's literally no base upstream the recognition site for SalI. Now, that you mention it. i can remember, reading somewhere that SalI needs something like 15-20 bases for corrrect cutting. However, the vector offers also a SacII and BgII site so either, I start designing new primers or - maybe the faster way - I ligate the purified PCR product into pGemT, cut it with NotI and SalI (the additionanl NotI-Site in pGemT shouldn't display any problems, right?) and then ligate in my original vector.

One more thing: If SalI didn't cut the fragment correctly, how is it possible that the fragment seems to appear in the vector later on (I "confirmed" this with a PCR on the miniprepped DNA)? The only possibility may be that parts of the vector only contain the NotI-digest but not the SalI-digest. Since this seems to be a rather rare event, that would explain the small number of colonies on the plates.

Exactly my idea. I digested each DNA sample with both REs separately and found out that both enzymes work fine, showing a band around 5.5 kb for every sample. Maybe, a sequential digest (although shouldn't be required) might help?