I was going to reply to Dimitry Libertas's post, as I believe mine is a similar problem to his, but I thought this would would have more emphasis if it was a separate post.
I am using utf8_unicode_ci collation and am very frustrated with the support for "expansions" (see
http://dev.mysql.com/doc/mysql/en/charset-unicode-sets.html). Evidently 'a-acute' or 'a-umlaut' is interpreted as "equal" to 'a' which is generally NOT what is desired. For example, a customer in Sweden is complaining that searching on man yields matches for Människan.
Using the binary collation (utf8_bin) is not a good option, as the sorting that this would produce is not desirable.
We could programmatically weed out false matches, but this would be disruptive and unnecessarily complicate our application which supports multiple database servers.
Also, as Mr. Libertas observes, the 'like' operator works differently than '='.