I am not able to comment on macrogen. As to Complete Genomics (CG), you should ask yourself about how many analyses you want to do by yourself. CG has a very sophisticated pipeline. If you want to leave your project to them, CG will be a decent choice. However, if you want to get their data and analyze by yourself, you should watch out. While you have a variety of choices of software for analyzing Illumina data, the CG software package is pretty much the only one that is able to process the CG reads and alignments. When you have troubles with their results, you can hardly get help elsewhere except CG (their support seems good; I have sent them a couple of requests and all got replies in hours).

Complete Genomics has made available data from 69 individuals. I picked up a copy on thumb drive at a recent conference but have yet to use those data. I know it is available from the web site. You could grab those data and take a look for yourself regarding data quality.

Check out the Assemblathon and ask those folks these same questions. While not directly related to your question, the Assemblathon was a test of different assembly algorithms but those involved surely would be able to comment on data quality.