short duplication of part of the insert ChIP-Seq

[Nihilo]

I am currently verifying the library quality of my ChIP-Seq preparation by sequencing some of the subcloned products into the TA cloning vector. Upon alignment of the insert (i.e. between the two Solexa primers), I frequently obtain a 80-90% alignment to the human genome with an extra bit of sequence that does not align. When I look at this extra sequence (usually between 4-9 nucleotides, but it can go up to 30 nucleotides), it is always complementary to the start of the insert. Here's an example:

...ttccgatctCCTTAATTxxxxxxxxxxxxxxxxxxxxxxxxxAATTAAGGagatcgg....
(end of Solexa primer) (rest of insert) (other Solexa primer)

The beginning of the insert (CCTTAATT) would align to the human genome with the rest of the insert up to the AATTAAGG sequence which would not align. In every case, this remaining sequence would be complementary to the beginning of the insert (in this case CCTTAATT). This occurs 70-80% of the time. On rare occasion, I will obtain an insert without that feature that will entirely align with the human genome. Is this normal ? What could be the cause of that ? It is always happening on the side of the Solexa PCR primer 1 (CAAGCAGAAGACGGCATACGAGCTCTTCCGATCT), regardless of the orientation of the insert in the TA cloning vector. Am I amplifying molecules that are circularized by a certain mechanism over other molecules ?

Thank you !