Hi Ian.
The reason for this is due to the inner workings of the Blackfin processor itself. Intuition would suggest exactly as you detail above, but the part you are missing is perhaps lost in context inside that info bullet...where it says:
"However, the internal voltage regulator controller will not function if VDDEXT < 2.5 V, and the RTC cannot be used to wake the core from the hibernate state.",
the key is the part that indicates the internal regulator controllerwill not function if VDDEXT < 2.5 V. The internal regulator was designed to work with a nominal 3.3V VDDEXT to buck down to the 1.2V domain. When the device was later extended to support 1.8V VDDEXT for mSDRAM, it came at the cost of no longer having the on-chip regulator to govern the VDDINT supply from the VDDEXT rail. Even though you are fully controlling these supply levels yourself in this scheme, the disconnect is that the regulator controller circuitry that remains enabled during hibernate to monitor the RTC wake-up event is not functional when VDDEXT doesn't get above 2.25V (this is the value in the datasheet in footnote 3 of the Operating Conditions table). The missing piece of information is that the RTC event itself is NOT generating the internal reset to recover from hibernate; rather, the event is sensed by the regulator controller (enabled by setting the WAKE bit in VR_CTL prior to going to hibernate), which THEN initiates the reset sequence. Failure to provide this option to sense the RTC event would otherwise force any RTC event to break hibernate, even if the user didn't want it to. So, in summary, even if the RTC is at 3.3V and generates the event, the regulator controller cannot sense it, so the RTC event cannot be used.
-Joe