I'm designing a gene in DNAWorks, and I'm unable to find a set of conditions that gives an "overall score" of zero (which is what I normally insist on). People that have used DNAWorks: what's the highest nonzero score you've been able to successfully assemble?
You might want to read Gibson's paper on the step-wise assembly of the mouse mitochondrial genome (1):
He started with 60b long oligos with 20b overlap, as he assembled 5 of those 60-mers into a backbone, obtaining 384b fragments. On the next step, he joined 5 of those 384mers, obtaining 1.2kb constructs. You can do the same, but on the second step use 2x 384mers to get you ~600bp gene. Gibson didn't use DNAWorks to chop up the sequence. He just started from base 1, so that his fragments were F1 [1:60], F2[41:100], F3[81:140], etc.
I think that 15b homology is pretty low (Gibson used 20b for assembling the 60mers in the first step, and 40bp homology for the next assemblies). 60b oligo length is standard.
You can also try one step assembly with all of your oligos (via PCA) as you planned, but I think that the two step assembly might be more efficient.
Let me know if you need more help.