I have no clue about C. elegans, but have done a lot of qPCR to do mtDNA copynumber determinations in human and mosue cells. As mtDNA exists in many copies per cell also in C. elegans, a single-copy gene will be difficult to get working in a duplex PCR with mitochondrial DNA as the second target. If the 18S gene is having a more or less stable repeat number in the genome and doesn't vary a lot between the strains/animals/conditions you want to compare you will have an easier life using this as reference gene than any single copy gene.

Thanks- I'm not actually doing any multiplexing- it's just a very simple SYBR deltadelta ct experiment. I have the reaction working very well with 18s but my comparison between the two is essentially meaningless because I have no idea what the copy number of 18s is in my species (I'm working in marine mammal parasitic nematodes, and there is almost no available sequence data). 

Assuming there is no good candidate reference gene ... and knowing that 18S is always like shooting fish in a barrel (although its mother samples have to be diluted ~1:1000 more than the same samples do for assessing most other targets), there is a semi-unorthodox way to look at this problem.  If, at ~100% efficiency, a single (single-stranded) reaction copy of target appears at ~38 cycles at a constant threshold dRn across plates (a Cq of 38 at 100% efiiciency for a single copy appears to be a very good estimate in most real situations with high-quality qPCR), then 2 single-stranded copies of a target (as in a dsDNA target) in a reaction would appear at 37 cycles, and so on.  Loading the exact same amount of starting material into every single reaction will allow you to glean copy # differences in the total # of mtgenomes you are probing, if indeed you know the number of mitochondria per worm, and you know that the copy number of your target of interest is indeed 1 per mtgenome.  The mtgenome DNA is double-stranded - don't forget. 1 strand is one physical copy, a double strand is 2 physical copies (albeit complements of one another; each is nonetheless engaged by either reverse or forward primer during the PCR/qPCR as an "event").  It would seem to me that the main factor here, is then, being very precise in your estimate of how many mitochondria/nematode?  The other most important thing is to have a very good estimate of amplification efficiency of your target of interest as this (of course) impacts directly on this semi-unorthodox approach (a stealth serial dilution study on a trustworthy representative sample is good for this; an extensive serial 1:2 dilution study from full-strength sample out to 15 points, with up to 8 technical replicates at each dilution point; then use EAMP = 10(-1/slope) to calculate your EAMP (qPCR exponential amplification efficiency value) from the most log-linear collection of the subpopulation of points that respond to dilution correctly; in log-linear fashion.  Also, realize, that when you are dealing with near single-copy reactions, a minimum of 16 technical replicate reactions are necessary to tease out the proper Poisson distribution analysis crucial to identifying such single-copy reactions in the first place.  E.g., "nil reaction" analysis...  Once you've identified (by nil reaction analysis) the Cq value at which 1 theoretic single-copy single-stranded target reaction reports (which I say is many times ~38), then, you merely plug in that discovered value as "b" in the formula:  initial # of copies in a rxn = EAMP(b-Cq) ...   This, in theory, allows you to estimate the # of initial copies in any reaction for that target generating a "Cq" in the equation.  And, this copy # can then be interpolated back to how many copies of the target existed per worm.  But, knowing the number of mitochondria per worm will be important in the end.