Is My Delay Compensation Totally Borked?!?

I know that this has been discussed before, but I have not found one thread that has solved timing issues with recorded audio in Sonar. I've come across several threads which describe similar problems, but they usual have resulted in long-winded explanations, arguments, and saber rattling of of egos, but no overall resolution.
 
The point of my post: is my delay compensation borked (i.e. defective, not working, screwed up, failed, etc.) or is something else going on here?!? 
 
1) I cannot for the life of me figure out how to get recorded audio to line up perfectly with MIDI notes. I don't want to have to nudge stuff around constantly.
2) Each time I hit record, delay/latencies are inconsistent. First there may be 6ms of delay, next may be 9ms of delay, etc.
3) Why does the External Insert plugin completely screw with latency? I mean, this thing sucks!

First, let me ask a basic question because I really don't know: if no matter what a plugin's latency may be, then shouldn't Sonar automatically compensate for it without having an effect of leading or lagging audio? For example, if I put an instance of the ProChannel Concrete Limiter on my master, it introduces more lag on recorded audio. Yes, I am aware that PCCL has a latency of 1.5ms, but should this not be accounted for and thereby not affect any recorded audio? In other words, it should be seamless/transparent to the user. Other plugins that I have tested don't do this (e.g. Ozone 5 with all modules turned on<--resource hog, or 6 instances of Valhalla Vintage Verb piled high). I find it odd that such a simple plugin such as PCCL can negatively tweak a project. It's almost as if PCCL is not reporting its latency to Sonar. Which brings me to another question: how can I tell if a plugin is not reporting its latency to Sonar? Is it basically the scenario I described?
 
I am currently in discussion with Access in regards to an issue I am experiencing with my Virus. Overall, if I record audio from this synth and another synth, which both go through the same audio interface, the Virus audio will lead by about 5-6msec the MIDI notes that trigger it.  The other hardware synth is normal.  Despite adjusting manual offset (+240 samples), I can get any other input other than the Virus to line up.  The more manual offset I add, however, the further forward the Virus audio gets pushed. It may have something to do with USB-"MIDI" triggering vs. traditional MIDI triggering.
 

 
And each time I hit record, the lead/lag may change slightly. Hitting record in Sonar never is consistent. Why could this be?
 
And finally, the External Insert plugin completely blows. If it is added to a project, then it completely and utterly wreaks havoc on timing. What's up with this thing?
 
I have done extensive, time-wasting testing on this. So far, I have tried the following, all of which have not improved anything:
 
1. Replaced nVidia graphics card with my old ATi graphics card.
2. Disabled and unplugged my MIDIsport interface and used an alternate MIDI port.
3. Disabled delay compensation for Virus Control.
4. Used the program LatencyMon to check for any outstanding latency issues among drivers.
 
I know I have asked many questions and have thrown a lot down on the table, but let's go with the easiest problem first: issues with the External Insert plugin.  This stupid plugin introduces HUGE and VAST amounts of delay for any recorded audio.  I mean this thing is abyssal!  It is so bad, it makes me think it's not even working right.  I mean, how the heck could a plugin deliver inconsistent latencies* and delays to recorded audio ranging from about 15ms to over 100ms!??  Who here uses it and does it give you problems like I am experiencing?
 
* "Inconsistent latencies" refers to a differing amount of delay each time I hit the record button.

Maybe it's not delay compensation that's the issue here, since AFAIK that applies specifically to audio being sent out and returned, not straight audio recording.
 
You could try correcting the timing error using the Manual Offset adjustment. Zoom in and set the time display to Samples, write down the value at the start of a quantized MIDI note, then check the sample number at the start of the recorded wave. Subtract the two and enter that value into the Manual Offset box. This will tell SONAR precisely where to place the recorded audio and it'll match the MIDI precisely.
 
The potential downside to this workaround is that if it's actually a MIDI timing problem, the offset correction could screw up non-MIDI audio recording. You can test that by recording a sharp sound such as a pencil tapped on the desk and see if that audio is now out of sync. If it's OK, or if you never use microphones, then problem solved.
 
But if the microphone audio is now whacked, then the manual offset isn't the right solution. Look at the IgnoreMidiInTimeStamps variable in aud.ini (try setting it to 1), and the "Prepare Using N Millisecond Buffers" in the MIDI options (try lowering it to 100ms).
 
(BTW, I have always had this problem recording outboard synths, a problem that has persisted across three different interfaces. I just routinely nudge audio to compensate. I manually insert a quantized MIDI note at the beginning for the sole purpose of lining up MIDI to recorded audio. I could use the manual offset trick, but then other recorded audio is off so I'd have to keep changing the value depending on what I'm recording.)

Be aware that when using different hardware synths they will each take a slightly different time to process the MIDI and get the noise out.

There's a two-four millisecond difference between my microQ and Mopho for example.

Different USB MIDI interfaces may also require different processing times in the order of a tiny number of milliseconds.

I routinely have to shift recorded hardware synth's audio a tiny amount to get things lined up correctly. What's more it's a different amount for each synth.

I use the external insert once in a while and usually have to tweak the recorded audio, again a tiny amount. However other than my interface I'm passing the audio through no digital convertors, just purely analogue gear. If the external stuff is digital that might be adding in latency of its own.

As for the difference between theoretical audio interface latency and "ping" real-time latency..... I'm using an iPad so can't see if your interface is in your sig or not. If it's a USB one in particular I'd hazard a guess that this may be down to "hidden" safety buffers in the interface, a not uncommon thing with USB soundcards. These buffers may not be reported to Sonar by the driver. Some interfaces seem to be very inconsistent between the driver-claimed buffer settings and what is actually going on.

I also wonder if you're looking at two different issues here. One being audio latency/sync and the other being the MIDI side of things.

^^^
I agree that it is most likely delay compensation of either Sonar or The Virus TI software that is putting the recorded audio ahead of the MIDI notes.  I have an open ticket with Access, and they are troubleshooting this for me.  They said they don't get a lot of Sonar questions much, so I think they went out and bought a copy to troubleshoot! 
 
Tonight, I plan on testing/recording the Virus sans the USB-MIDI connection; I'll just use a plain old MIDI cable and trigger it that way.  I suspect timing of the recorded audio will be on par with the Waldorf Q and not in front of the MIDI note.  I'll post my findings.
 
As far as the External Insert plugin...I don't know what to make of that thing.  I mean it works but it is extremely flakey.  It offsets recorded audio, and it sometimes offsets live-played audio, and it sometimes totally misaligns the metronome and/or a host-sync'ed drum synth. It is completely unstable and almost unusable.  I guess I'll need to go back to the old-school method of setting up an aux channel-return instead of having a nice, all-in-one bus.