Jim Roseberry
That's a good tip.  
The added latency on the reverb is more pre-delay.

Exactly.

vintagevibe
Another trick is if you you have a direct monitor mode on your interface but no DSP, add delay/verb SEND on your recorder channel but have the channel itself muted during recording.   That way you can have zero latency feel plus FX and any latency in your verb/delay send will just make it sound like a bigger room.

Yeah, I'd been doing that to fake a better feel. But it still wasn't spot on in timing. Had the reverb, which sort of helped to disguise the latency.

With the mixer method I've mentioned for my scenario, now It's spot on. It's a huge win for me.

I'm glad MOTU's interfaces have on board dsp so I can add compression and or reverb on the interface with it's zero-latency CueMixFX interface.

I was never happy with using the interface monitor system.. There was never any latency as you are hearing you voice or guitar going in directly while it's still analog.  If DSP effects are involved then yes the signal has now passed into the digital side so A/DD - D/A latency is now added. 
But my reasons for using a very similar system as the OP is headphone and control room level issues. And the option to add reverb to the headphpones. 
My Mike or a direct input for bass or accoustic guitar go first to a Joe Meek 3Q. 
It has a toggle for mike / line so that solves that issue. You can leave them both plugged in. 
The Joe meek has parallel outputs so one goes to a Yamaha MG 82 CX mixer and the other to my Focusrite 6i6 interface. 
The 6i6 1/2 outs go to the studio monitors and 3/4 go to the little mixer. 
The mixer gives you a solid clean mix that's easy to blend for either me or clients. 
If its live tracking of vox and acoustic guitar,, I can add the guitar to the mixer using it's line out or a second mike, lots of quick easy solutions. 
I also have a Mackie mix 8 that cost $70 that I can add in there if there needs to be more headphone mixes. 
It's funny the 6i6 has 2 headphone channels and powerful mixing software but the hardware is much easier to use and always works. I have no love for Mix Control and it illogical set up. 

I've been direct monitoring for years.  Much less headache.
 
I even record MIDI without monitoring through the computer.  just record the MIDI while monitoring the keyboard audio.  it doesn't have to be the exact sound you end up using in the project, just something close.  for example, if you were going to render your MIDI to some strings in Kontakt, for example, just record the MIDI but monitor the keyboard using string sounds from it.  then render the MIDI to kontakt after editing the MIDI.  no need to even have Kontakt inserted in the project until the MIDI has been recorded and edited.
 
and as gustabo says, my MOTU has dsp I can add through the cuemix software, so no need for external sends to verb or compression.

Direct monitoring is what I resorted to as well. The latency through the DAW is not reliable at all when using even a single low-latency compressor plug-in. However, I need to use a light compressor for vocals being tracked. Even freezing all the Sonar tracks doesn't provide a low-latency tracking ability, something I'd really like to see soon. Coulda saved me a few hundred $$. Even MOTU could have saved me this money if they offered plug-in support for CueMixDSP using my MOTU 24CoreI/O-PCIe.
 
My solution
I use a DBX560a (http://www.sweetwater.com/store/detail/560a) 500-series module next to my other 500-series modules. I use a Radial Workhorse Cube: http://www.radialeng.com/cube.php
Pre's I use (in the 500-Cube or not) can be split off via XLR and 1/4" from this unit's rear and the 1/4" can be fed into the input of the DBX 560a. The output of that can be fed into the head-phone amp mixer. You don't need the 500-Series stuff to accomplish this. You just need a compressor and the ability to split off your preamp's output to both 1/4" and XLR without signal degradation (not typically a problem in decent mic pre's).
 
Thought
What I find interesting is, go back to the year 2000 and check out a Yamaha AW4416. This unit offered DSP hardware that allowed you no-latency tracking with compression on the listen-back head-phone mix. We still haven't been able to do this well in today's hardware/OS. 15 years later.

this is the compressor I use on the way in (which I believe is the same compressor br is using but in a rack model instead of a 500 series):
http://www.sweetwater.com/store/detail/160A
 
for pres I use either an ART MPA Pro II http://www.sweetwater.com/store/detail/ProMPA2 for color
or a True Systems P-solo for clean mic pre http://www.sweetwater.com/store/detail/PSolo
 
I use a Mackie for bringing everything together for monitoring and I use the mackie "split" inserts for keyboards.  Bass and Guitar, if recording direct, go into the Mackie as well, but typically are mic'd.
 
from the Mackie, I send the mains out to the near fields but for recording, I use the ALT 3/4 outs going to headphone amps.

brconflict
 
Thought
What I find interesting is, go back to the year 2000 and check out a Yamaha AW4416. This unit offered DSP hardware that allowed you no-latency tracking with compression on the listen-back head-phone mix. We still haven't been able to do this well in today's hardware/OS. 15 years later.

yes, i had this same availability with both my roland VS-880ex, and my Roland VS-1880
 
latency was never even discussed at the vs planet forum, a non issue.

batsbrew

brconflict
 
Thought
What I find interesting is, go back to the year 2000 and check out a Yamaha AW4416. This unit offered DSP hardware that allowed you no-latency tracking with compression on the listen-back head-phone mix. We still haven't been able to do this well in today's hardware/OS. 15 years later.

yes, i had this same availability with both my roland VS-880ex, and my Roland VS-1880
 
latency was never even discussed at the vs planet forum, a non issue.

Obviously, this is now caused more by latency-inducing plug-ins, which are far and wider more CPU-intensive than ever before, especially when we're talking 'moddled', but I haven't found any included plug-ins that have low enough latency-induction to be real useful. More, in my case it was a problem that sometimes the latency is high in the Echo function, other times, it's near-zero. It's never been consistent unless you disable PDC.
 

Glad others see the need and reasonableness. It works for me!