Ligation artefacts

[Astala Vista]

We recently constructed a small human genomic DNA library in a plasmid 
vector for use in an undergraduate practical class. The human DNA was 
partially digested with Sau3AI and was size selected for fragments 
between 2 kb and 4 kb. The plasmid was pUC18NotI which has NotI sites 
flanking the normal pUC18 MCS. The vector was digested to completeion 
with BamHI and was then phosphatased. Ligations were set up and DNA was 
electroporated into E. coli DH5alphaE. Colonies were picked at random 
and mini-prepped. Around 100 random clones were sequenced from each end 
of the insert using primers located far enough back that we could read 
through the plasmid/insert junction.

As you would expect, a minority of junctions recreated the BamHI site, 
and the majority only created a Sau3AI site. However, 7 of the junctions 
had neither a BamHI site or a Sau3AI site. Several bases had been 
altered in the vicinity of the jucntion.

Has anybody seen such a high level of ligation problems before? Any 
guesses about what's going on?

Thanks.