I will refer you to following diagram that should explain how paired-end reads provide information about how long the fragment is that is being sequenced when a reference is available for mapping (Image comes from A: What is the different between Read and Fragment in RNA-seq? ). Two reads coming from the fragment being sequenced would align much farther apart than just the size of the insert, if the fragment contains a splice site.
Single-end reads would be equivalent to
Read 1 or
Read 2 on its own without the mate.