I have a recent post on this subject HERE. In brief:
The online & downloadable NHTSA FARS data for 2009 data show the following for the posted speed limit where fatal bicyclist-motorvehicle crashes occurred (I graphed this before the 2010 FARS with the PBCAT data was available):

For those unfamiliar with cumulative quantile plots such as above, the y-axis is the cumulative percent of dead bicyclists at or below the posted speed limit on the x-axis. Each point on the graph represents one dead cyclist. With hundreds of dead cyclists/year, each dot is quite small and they then merge to form a "line", but if the original graph is viewed at enough magnification you can see the individual data points.

Reading the quantile plot shows, for example, that only about 20% of fatal crashes are on roads with speed limit of 30 mph or less. Not surprisingly, higher speeds are associated with most of the fatalities. My estimate is more that 20% of bicyclist exposure (time or distance traveled) is on roads with posted speed limits of 30 mph or less, so the fact that only 20% of fatalities were on such roads indicates they are safer.