Failing to Experiment and Test
Testing and experimenting are absolutely vital if you want to achieve maximum CTR. Although some people do get lucky and reach a very good CTR with the first version they try, most people will need to test a few different varieties before they find a version that works. Different demographics will react in different ways to different styles. Not only that, but different groups of traffic within that demographic may also react in different ways. For example, you might find that on a site with a majority of its traffic from pay-per-click marketing, you get great results when you've blended the ads into your website. But for traffic from Digg, that same strategy might not work.
This means you may see sudden drops or spikes in CTR when you start receiving a significant amount of traffic from a new source. I'd suggest starting out by blending your ads into your website. This seems to be the strategy that works best for most people. Test until you've received a decent amount of traffic and clicks. You can't get an accurate test with only 100 impressions. Ideally you'll want to wait until you've had at least 1,000 impressions before you make a judgment about the effectiveness of your ads. So what is a "good" CTR?
Well, honestly that depends on the niche, the traffic sources, the demographic, etc. There are many factors that can affect CTR. People say double digit CTR is good. Under double digits is still ok, but tweaking can very often bring CTR into the double digits. Double digits over 20% are possible, but this can't often be achieved with extensive testing. What can be tested?

• The position of the ads.
• The size of the ads.
• The background color.
• Blending vs. standing out.
• Different colors for links, text, border.
• Different border styles (none, standard, rounded).