The paper that phil linked to (thanks) discusses the detection threshold and shows a number of error sources (for example, high-energy cosmic 'rays" causing ccd artefacts). As a result of these various sources of error, they now estimate about half the detection rate as expected pre-launch.

This is still pretty good. 40-50 good candidates in a 60-day observation is excellent - enough to play statistics with the distribution of orbit frequency vs transit "depth" and stellar magnitude.

One interesting result that is NOT discussed in the paper:-

They list 145 eclipsing binary star systems. This figure can be compared to the ~50 candidate transiting planets, so we finally have a handle on the distribution of planets and their probability.

close-in planets with orbital periods of up to 30 days are 1/3 as common as binary star systems with the same period.

Now, some back of the envelope (and very suspect mathematics) - about half of all stars are in binary systems (or in other words, 1/3 of all "stars" are actually binary pairs), so this implies that about 1/6 non-binary stars may have a planetary system. (I'm grossly extending the results from short-periods to all periods here!)

These numbers need careful scrutiny, and the range of error is wide, but it's showing that many stars will have planetary systems - more than a few percent, but perhaps less than half?