Discrepancy in paired-end Illumina data

[kopardev]

I have a variety of paired end data sets from Illumina and these are some of their sequence lengths:
Read1 Read2
103 110
103 107
101 101
98 107
Can anyone explain to me, why read1 and read2 have different sequence lengths for some datasets?
Thanks,
Vishal

[simonandrews]

It's pretty unusual to get this from a standard Illumina run (you can make up a custom recipe to do it though).

Could the samples have had custom barcodes on the front of the first read, which have been stripped before returning the sequence to you?