I've written a lot about my skepticism about evolution, and here's a post on The Loom about mutations in the HIV virus that will serve as an illustration of of one of the assumptions that evolutionists make without, in my opinion, much merit: that genetic similarity implies common ancestry. Carl Zimmer -- who undoubtedly knows more about biological evolution than I do, since I'm a computer scientist who only simulates evolution -- displays and explains several charts that show how various strands of HIV are related to each other and to SIV (Simian Immunodeficiency Virus).
I have no doubt that human DNA is similar to chimp DNA, and that what Mr. Zimmer calls "microevolution" (a term originally created by skeptics of evolution?) occurs. In small time scales, changes can be traced between SIV and HIV, and their relationship can be observed and tested. However, the "wall" that Mr. Zimmer refers to is that of time, and it's a fact that any common ancestry between chimps and humans can only be inferred, not proven or even reproduced in a scientific manner (without a time machine). There's a gaping chasm between demonstrating genetic similarity and proving common ancestry that science cannot hope to bridge. The only difference between those who believe that God created life and those who believe it evolved is that the latter tend to fill in the holes with deception rather than faith.
Random mutation and natural selection are real and can be observed, but there's very little evidence to suggest that they are capable of explaining the origin of life. Predictions based on evolution are often very wrong. Just like everyone else, scientists have agendas and research tends to follow the grant money. (All that means is that you have to abandon the illusion of scientific objectivity and be prepared to consider the motivations and prejudices behind everything that's published (including this site, of course).)
Finally, I'll add a note from my own research and simulations. In my experience, similarity seems to be the rule for evolutionary systems. I've run countless simulations with non-interbreeding species with randomly generated genomes in NS+M environments (natural selection plus mutation), and their genomes almost always converge. Why? Because evolution is an optimization algorithm, and if it's implemented properly it will find an optimal solution to whatever evaluation criteria is applied to it. There are often local maximums that can trap poor implementations (perhaps analogous to "niches" in a biosphere), but over time these get shaken out and the result is uniformity, not diversity.
Phelps comments and posts and link to an fascinating story about plants that defy normal laws of heredity.
Challenging a scientific law of inheritance that has stood for 150 years, scientists say plants sometimes select better bits of DNA in order to develop normally even when their predecessors carried genetic flaws. ...
In the Purdue experiment, researchers found that a watercress plant sometimes corrects the genetic code it inherited from its flawed parents and grows normally like its grandparents and other ancestors.
There's a lot we don't understand yet, and much of what we think we understand is probably wrong.