basic Q about paired end reads from HiSeq

[panther]

Hi, I'm new to genomic data and I'm just learning the basics of Illumina HiSeq and the resulting data. There's one thing I'm confused about.

With paired end reads that don't overlap, do we know the exact insert length without mapping to a reference? Or is it just an estimate based on the distribution of fragment lengths we expect from our library prep protocol?

I've seen all these posts and videos talking about the insert size of paired end reads, but I couldn't figure out if there's a step in the sequencing that also keeps track of the overall fragment length, which would allow us to know the insert size by subtracting off the length of the two reads (and adapters).

Thanks!

[luc]

With paired end reads that don't overlap, do we know the exact insert length without mapping to a reference? Or is it just an estimate based on the distribution of fragment lengths we expect from our library prep protocol?
No and then yes.
Also note that shorter library molecules tend to cluster better than longer molecules.