Looks like the postdoc's hypothesis is correct

Here are the results from my quick-and-dirty test of how phage recombination happens in competent H. influenzae cells.  They show that pretreating lysates with DNase I reduces recombination, as expected if recombination happens between replicating phage and free DNA brought in from the lysate by the competence machinery.

(I'm an idiot.  Why didn't I do an infection with both lysates treated with DNase I?  Treating only one lysate is only expected to eliminate half of the potential recombination under this hypothesis, but treating them both would eliminate it all!  Next time...)

(I forgot to label the Y-axis - it's the recombination frequency.)  A is the recombinant frequency seen when cells were infected with untreated lysates of phage mutants ts1 and ts3.  In infections B and D, one or the other lysate was pretreated with DNase I, and in infections C and E, one or the other lysate was pretreated with Proteinase K.  The blue bars are the fractions of infected cells that produced wildtype (not ts) recombinant phage, and the red bars are the fraction of the phage output that were wildtype.

The data are not very accurate because I titered the cells and lysates by spotting dilutions on lawns and didn't try to count the phage in spots with more than 50 plaques.  I'll repeat the experiment with more careful titers.

I also titered the input ts lysates after their pretreatments.  The proteinase-treated lysates had 3-5-fold fewer plaque-forming units than their DNase-treated counterparts, suggesting that proteinase digestion did reduce infectivity but not recombination.  I should have also retitered the untreated lysates but I forgot to; the titers of the DNase-treated lysates were about 1/3 lower than the previously determined titers of these lysates.