I digested two plasmids, one with EcoRI and AgeI and the other with EcoRI and XmaI. Digests looked as expected, so I purified the respective fragments and set up the T4 DNA ligation (AgeI and XmaI sites are compatible).
After E.coli transformation, culture and miniprep from 12 of the resulting cultures, I digested the isolate from the minipreps with BclI to confirm I had the correct product (because there was one BclI site in the insert and another one in 1.4kb away in the backbone). The results of this digest clearly show that there are two different ligation products: four lanes share one banding pattern and the other eight share a pattern which is different from it.
I checked both plasmids for other restriction sites of the enzymes used which I overlooked before, but found none. Because EcoRI and XmaI/AgeI sites are radically different (AATT vs. CCGG), I would exclude the possibility of the insert being ligated in reverse orientation.
Are there any common causes of unexpected ligation variants besides unnoticed restriction sites and reverse ligation?