6. Recovery
Put the pad under the dissecting scope (on the inverted plate lid) and using a P200 pipetteman place a drop of recovery buffer on the oil drop above the worm. Then poke a worm pick straight down through the recovery buffer and oil to touch the agarose pad next to the worm. This will form a channel, and the recovery buffer will form a layer underneath the oil in which the worm will float. Worms can be left on the pad in recovery buffer for hours, but you might as well immediately move them to plates. (Some say it is better to leave them in recovery buffer for > 5 minutes - in this case place the coverslip in the lid of a large worm plate, and place the plate over it to make a humidified chamber.) Can put up to 3 injected worms on a plate; I prefer one worm/plate. Use a slightly drawn out and broken off and flamed smooth large diameter micropipette (1.5 mm diameter drawn out to about half that) and mouth pipetted the worms over to a plate, and set a 20·.

7. Results
Three days after injection, score the F1 for the marker gene phenotype (e.g. rollers if pRF4 (rol-6) is used). rol-6 animals are Rol even as young larvae, so it is tempting to score and pick the F1 after only two days: don't do this! The young larvae are very delicate and you are liable to kill them by picking them. Each Rol F1 is considered an independent transformant (even if several come from the same injected P0). Therefore, each Rol F1 should be placed on a separate plate to try to get lines.
Typically people inject 30 P0s (takes just 2 hours if you're good), and expect to get 3-300 Rol F1. Usually some of the injected P0s give zero or 1 Rol F1, most of the P0s give 5-15 rollers. As a beginner I averaged 1-2 F1 rollers per P0. Now I average about 8 per P0, and some people do much better. Of the Rol F1, typically about 5-30% will transmit the array, allowing a line to be established. Typically, lines transmit the array to 30-80% of their progeny. There is variation among lines transformed with the same DNA. For example, only a fraction of lines transformed with a cosmid/pRF4 might give rescue of a mutation in a gene found within that cosmid, and the strength of the rescue will vary among lines that do show rescue. In the Horvitz lab, people look at ~6 lines before they tentatively believe a negative result.
Some lines transmit at only a few percent per generation. The frequency of transmission varies from animal to animal. Jin says to be sure to keep these lines when injecting ß-gal or GFP constructs; the low transmission rate of the extrachromosomal array is useful when trying to select for chromosomal integration of the array (leading to 100% transmission).
Some people (me, Mark , Gillian, Jin, Tory) have noted a high incidence of males in the F1 of injected animals, or in some Rol lines. Since this was observed using a variety of different DNAs it is likely a nonspecific effect of extrachromosomal DNA on chromosome disjunction, and doesn't mean your gene is involved in sex determination.
If you are rescuing a mutant, and using pRF4 as a coinjection marker, you may notice an odd effect; a high proportion (up to half) of non-Rol F1 progeny of rescued Rol animals may themselves also be rescued for the mutant phenotype. This is not necessarily indicative of maternal effect rescue of your mutant. Rather, it can be due to lack of penetrance of the rol-6 dominant allele and/or mosaicism for the extrachromasomal array. This can be demonstrated by picking individual non-Rol rescued animals and showing that they throw Rol progeny.
8. Alternate coinjection markers
Sometimes it is not desirable to have the dominant Rol phenotype in your transgenic worms. In these cases, you can microinject worms carrying a recessive marker mutation with a rescuing plasmid for that gene, along with whatever your experimental DNA is. The following properties are desirable for such a marker mutant: 1) it should be very easy to score, preferably at all stages of development. 2) the mutants should have healthy gonads that are easy to inject. This is sometimes achieved by using a ts allele, growing the animals at the permissive temperature before injection, and then shifting to the nonpermissive temperature. 3) it should be possible to get strong F1 rescue of the mutant. 4) the rescued animals should be truly wild type.

I've heard about people using unc-76, dpy-20, and lin-15 as coinjection markers. I've been using lin-15. Its main drawback is that it can only be scored in adults. lin-15(n765ts) animals are raised at 15· for injection. A lin-15 rescuing plasmid is included at 50 ng/µl in the injection mix. I'm using the plasmid pL15EK, which I got from Xiaowei Lu. This is an 11 kb Eag1/Nru1 rescuing fragment of cosmid C29B12 cloned into pBSKS+ cut with Eag1/Kpn1, (using a Kpn1 linker on the Nru1 end). After injection, the worms are moved to 20· or 25·. At 25· the non-rescued animals are very sick, and there is a strong selection for transgenic worms. At 20· the worms are healthier, and the transgenic worms are recognized as non-Muv. You should wait 4 days after injection at 20· to score the adult F1. Even though the Muv phenotype only develops during the L4, it appears that at 25·, n765 animals reach the L4 stage more slowly maternal rescue. People usually put lin-15(n765)/+ animals at 22.5· in order to enhance the Muv phenotype of the n765 homozygotes they throw. In two trials using lin-15 as a coinjection marker, I got about the same number of F1 non-Muv animals as I usually get F1 rollers using rol-6 (i.e. about 100 F1's from 20 injected P 0 s). However only 4% and 8% of these F1s transmitted their array, whereas typically more than 20% of my Rol F1s transmit. Piali in the Bargmann lab says she gets about 15% transmission; she's using a different lin-15 plasmid than I am.
I find that lin-15 is much preferable to rol-6 as a marker when trying to integrate an array. The Muv phenotype is incredibly easy to spot, whereas screening plates for the absence of non-Rol animals (which you do when trying to integrate pRF4) is a lot harder.