Hello,
I ran a timing test today using Sonar to see if indeed it is keeping perfect timing.  (Yes, I am looking for bang-on-the-beat timing - needs to be perfect).
 
So what I did was I loaded up a softsynth VST and dialed in a closed hat sound with immediate attack.  I used the step sequencer to draw in 16th notes.   The notes are perfectly entered at each 16th tick:
 
1:01:000
1:01:024
1:01:048
1:01:072
and so forth...
 
(by the way, I am using a TPQN setting of 96, hence the numbering system you see above)
 
So when I print my audio, I would expect to use the "Tab To Transient" feature and expect the tabbing to go to the values listed above.... but they are ever so slightly off:
 
See video here:
 
https://youtu.be/WBHATXPxnLw
 
I would like to understand what I can do about this.   I know to some this might seem silly, but to me it matters.   My CV/gate stuff has no mercy, and I would like Sonar to be the same with softsynths.
 
Thank You!
 

Which softsynth?

i tried several as a test.  
 
the best test would be Battery with a sample that starts exactly at 0 with hard attack.
 
just ever so slightly off, the the biggest of deals but just wondering.

Yes just wondering because virtual analog synths might have some variance in the envelopes on purpose and a drum specific synth like AD2 uses round robin for some samples which could also account for variance. A sample played at velocity 127 in Battery with an attack of 0 and no randomization or humanizing should be perfectly consistent.

Tab-to-transient depends on a transient detection algorithm. It may not see the absolute start of a sound, but maybe a peak or some other threshold detection. It would be a good idea to examine the waveform to see where the sound really starts.
 
As others mentioned, the exact envelope and/or sample performed by the synth could account for late-sounding notes. Otherwise, there would be no excuse or explanation for MIDI that is all "in the box" to be detectably off. It's another thing altogether to transmit MIDI out an interface and back in to trigger a synth (external or softsynth).

OK thank you for the suggestions.  I think there might not be much of a problem here.  I edited the sample in battery to start in the middle of the waveform, and now its spot on (except for one hit out of like 200 - but thats ok)... I think my issue was the slightest bit of an attack problem.

Good to know though!  Thank you for the help!