I think I may have made an important discovery regarding GPO legato mode and latency settings in sequencers. Please note that I have not rigorously tested this, but early research suggests the below findings. I am using Sonar 5 Producer Edition, but I strongly suspect this will affect users of any sequencer.
I at one point noticed that at very high latency settings, the legato function seems to not work. I am talking in the neighborhood of 150ms+. I rarely work with such high latencies, but if I'm editing a huge score and haven't frozen many tracks, I might bump it up temporarily. But now I think I've found a direct correlation between the latency setting and how legato functions.
It seems that the time between a sustain-on message (ctrl#64=127) and the beginning of a note must be greater than the effective latency for legato to work.
Let's take an example: Suppose you have a piece whose tempo is 120 bpm and you have a legato run of sixteenth notes. There will be about 125 ms between the beginning of each note. If you place a ctrl#64 message exactly between the first and second note, you have about a 62 ms "clearance" before the second note sounds. In this example, any latency setting above 62 will cause the legato to not function for the second note (though it seems to kick in for the third, if legato is held).
Now 62 ms is pretty high, and most use a much lower setting. However, if the message had been placed closer to the second note, or if the tempo were faster, or if 32nd notes were used, your clearance would become much smaller. For very quick alternating tongued and slurred notes in a woodwind instrument, you will have to be very careful about controller placement, or the slurred notes will break apart, or even be lost all together.
This problem can definitely be worked around, so I thought I'd spare folks a little grief if they are having trouble with legato. I'd be curious if anyone has anything to add to this experiment, conflicting or supporting.