Another possible solution to delaying the turn on of the +5 VRef supply is to increase the capacitance load it sees.  The delay waveforms show two distinct slopes.  Initially it charges at very low rate.  This is because the cap held the voltage low long enough so the fold-back circuit turned off the boost transistor.  During this period of time it charges at the LT1021 current limit rate.  As soon as the voltage gets high enough and turns off the fold-back transistor then it charges at the current limit rate of the boost transistor which of coarse is much faster.  This fold-back is what makes it a little strange.  If you use a small cap, say 22uf, you do not get half the delay as the 47uf does.  In fact it does virtually nothing.  As all the supplies are coming up and the devices are getting potential to work a smaller cap gets charged too fast and the fold-back doesn't kick in so the filter cap is charged at a 200ma rate so there is very little affect on the turn on time.  Get just large enough cap to activate the fold back and it really stretches it out.  If the fold-back wasn't there would need a cap in the multi hundred uf range to slow things done enough.      Another thing to watch out for is as the output cap gets very large then when the power is removed from the unit the turn-off times become a problem.  Look at the 220uf turn-off plot below and you will see the the +5VRef is a volt higher then +5VF again.