CIAF should also be written.

Small events should use an approach similar to CLEVER, but use an emperical model of the read pair distribution to estimate the RP support breakdown between REFPAIR and DP.