A few things to consider:
- If you are finding wild-type colonies, it indicates your DpnI is not working optimally or you have too much template included. I've found that with good DpnI, a 2 hour digest easily eliminates 200+ ng of template and produces no background colonies (I add 10-20 units of NEB DpnI per 50uL SDM reaction for 50-250ng of template, respectively). Make sure to include a polymerase-free control reaction to establish your background.
- Sometimes the primers or a given annealing site in the template are not efficient. Do you have any predicted secondary structures in your primers, or low-complexity/high GC regions that might compete for annealing? These types of problems can ostensibly reduce the number of correct, complete DpnI-immune molecules, and hence deliver no mutant colonies. You can use DMSO in SDM (2-4% final concentration in my experience) to deal with difficult primer/template combinations.
- The purity of primers can be a concern, particularly if the mutation is located toward the 5' end of a "forward" primer. Standard desalted oligos might contain heterogeneous populations of oligonucleotides, including truncated primers that lack the full 5' sequence or contain incorrect bases (see a brief explanation here). You mentioned an irregularity in the primer design - might it have caused this? I play it safe and use PAGE purified primers in most cases, but others find success with standard desalted oligos (they are cheaper to reorder if you run into problems, at least).
- Increasing the extension time for troublesome SDMs can help. Sometimes doubling the recommended extension time makes all the difference. Even with the faster polymerases like Phusion and Q5, I find that 1.5-to-2-fold increase in extension time gives me mutants, whereas the recomended times do not. I rationalize this by considering that plasmids are large molecules, and polymerases don't always hold on for the entire length or start amplifying right away in a given amplification step. Hence a little extra time provides more opportunities to complete the job.
- Increasing the amount/purity of template (minding the need for good DpnI, as stated above) can sometimes solve SDM problems, too. The amplification in Quikchange-style SDM is not exponential - it proceeds linearly - so an increase in template directly increases the amount of product possible, assuming primers are present in excess. Usually, 125ng of each primer is used with this method, and the plasmid is often two orders of magnitude more massive (1000's vs. 10's of nucleotides in length). So using more than the recommended 25-50ng of template won't limit your reaction by exhausting your primers.
- How much of your transformation are you plating? Maybe try plating the entire contents of NYZ/SOC outgrowth to assure any mutant clones make it onto the selective plate. Again, make sure your DpnI digest works well before doing this to avoid increasing background colonies.
Good luck with your experiments.