Exactly what are you trying to do?  If you are trying to model a 3-hour event, your stormwater regulations should provide the necessary rainfall distribution.  Anything other then standard 24-hour SCS/NRCS rainfalls is basically a custom distribution, and should be modeled using the specified rainfall table.  For example, Autstin, TX has a complete set of 3-hour rainfall distributions.

When you COMPRESS a 24-hour Type III storm to 3 hours, you increase the rainfall intensity 8-fold.  (Assuming the same total depth.)  All intensities, including the peak, will be eight times greater.  This is not an intended use for this rainfall, and the results are not valid.  Only a few specific storms (such as the Illinois Huff distributions) are intended to be used at varying durations.  In general, if a storm is designated x-hours, it should be used only for that duration.

On the other hand, if you EXTRACT the middle 3 hours (around the peak) you will accurately preserve the intensity information represented by the original curve.  Of course, you will need to use the truncated curve with the appropriate 3-hour rainfall depth for your location.

You are correct that a 5.7" rainfall over 3 hours will be much more than a 25-year event.  To determine the corresponding return period you would need a table of depth vs. return period for a 3-hour event.

I ask because a colleague trying to find fault with the design of a dry swale has used a hydrograph (one that we used to determine peak flow) to calculate volume and say that our dry swale is undersized. I don't think the hydrograph should be used for volume, especially since it compresses the majority of the 5.7 inches of the 25 year type III 24 hour storm into a 3 hour period. Then he says that the municipal storm drain will surcharge during this three hour period, activating the valve that prevents backflow into the swale, but which also prevents the site water from draining so the water will build up in the swale and spill over the emergency spillway. I can tweak my design to accommodate this higher volume number but I don't think it should be labeled a 25 year storm event because the rainfall is so compressed. I would never size a swale using a hydrograph for volume, so how do I defend against that kind of attack?  

To get the rainfall data for any duration and frequency  go here:

http://www.weather.gov/oh/hdsc/

From the PFDS I found Technical Paper 40, which has a map showing around 3" for the 3-hr, 25-yr storm for CT.  The 3-hr, 100-year map shows CT smack in the 4" range.  This being said, it would follow that 5.7" in 3 hours is a greater than 100 year event.

Y'all are awesome. Thanks.

To do a 3-hr hydrograph use the local IDF curves that plot average rainfall intensity vs duration of the raincloud.
Volume = raincloud duration x average intensity.
You can do a several durations to develop the hydrograph.
For the 3-hr base use 1.6 hrs for the leading edge, 1.4 hrs for the tailing edge, or whatever seems reasonable.
Rainclouds are huge -- a 5-minute duration raincloud moving at 12 mph is one mile long!!
Your site is probably a lot smaller so use the 5-minute duration for all your runoff calcs.