Help me understand interfaces, RTL, and the MIDI controller

I am posting this in the Getting Started" forum because I am sure that other people may have these same questions.
 
I will make a few assumptions that I believe to be true but I want to be sure I understand so that I don't waste money on hardware that I do not need or won't do what I want.
OK, as far as having a fast RTL (Round Trip Latency) I assume this does not matter it I am recording one track at a time and that the track is not using effects or software programs that are on the computer or in the DAW. For example, say I have a POD or effects unit/pedal or DI from an guitar amp and have it running into my interface to record a guitar track. In this case I can use "direct monitoring" and in that case RTL is of no concern to me... or say I am recording a dry vocal track with a mic connected to my interface. I can use "direct monitoring: with no concern about latency. Is that correct.
 
Now, if I want to record a guitar track using Guitar Rig 4 or Amp Farm or any other DAW based software, then I have to consider RTL and cannot use "direct monitoring"? Is that correct?
 
Now, say I have a USB powered MIDI keyboard controller such as the Cakewalk A-800 Pro. It connects to the computer via USB and is USB bus powered, however... I am confused as to what would be the best way to connect such a controller. Say I buy a Roland Octa-capture which has great preamps and a pretty fast RTL. Would it be better to connect the 800 thrugh the Octa or just connect it ia USB? Which would have the lowest latency? IS this a concern.
 
In anotheeer thread Zo said (about the Octa-capture compared to other interfaces in this price range) quote:
Forget about anything else in this price range .......
Go octacapture , or Rme babyface or firfece if you got the budget ....
The rest is just not in the same league .......including motu !!!
We made a test of some crazy converters and just for fun we tested the octacapture and we were like dammm !!!
For the price it's unbelievable ....
 
Then
 
Then Jim Roseberry posted this, quote:
The OctaCapture yields 7.2ms RTL at the 48-sample ASIO buffer size/44.1k
 
The MOTU pre mkIII units yield 5.5ms RTL at a 64-sample ASIO buffer size/44.1k
The mkIII units are right about equal to the OctaCapture RTL wise (albeit at a 64-sample ASIO buffer size)
 
The RME Babyface yields 4.9ms RTL at a 48-sample ASIO buffer size
 
The M-Audio FastTrack Ultra 8R yields 5.5ms RTL at a 64-sample ASIO buffer size/44.1k (in high-performance mode)
 
FWIW, I've had good luck with all the above...
 
I'm not a fan of typical USB/Firewire units... as they use a large/hidden safety buffer... which results in huge RTL (making effective software based monitoring impractical - ie: Playing/monitoring in realtime thru an AmpSim plugin)  
Whatever the OP chooses, I'd recommend going with a proven rock-solid unit.
After the PC itself, that's the second most critical piece in achieving a rock-solid DAW.
 
So, should if I am going to use programs like Guitar Rig when recordinga guitar part I need to consider RTL. Correct? If I am using DAW effects when recording a vocaltrackI need to consider RTL. IF I am recording dry or with a POD or other preprocessed signal, RTL is not a concern if I have direct monitoring, correct?
Now considering all this... if I want to use a MIDI keyboard controller to play and record a softsynth, then RTL is a primary concern. Correct?
I assume that the quality of the mic pres are also a primary concern. So, if I were to buy one of the interfaces in Jim's list, I am leaning toward the Octa but some of the lower RTLs listed make me wonder which would best serve me (I also assume that the ability to work well with X1 is a big plus and being able to use ACT and have a good HW controller would be a plus).
 
I a handful of mics and most of my recording will be using two mics, a phantom powered condensor and a dynamic when recording a scratch track. Then if I do vocals and acoustic guitar at the same time I will usually use one powered mic and two dynamic mics... occasionally four mics. I am pretty sure I would always be usingthe Midi keyboard controller alone. Still I don't know if it would be better to get a plain jane keyboard controller and let it use the interface driver or get a better controller like the A-800 Pro and run it uisng the USBconnection (or through the interface).
 
So... I hope my questions are not too confusing but I want to buy what I need and what will work best for me. I have wasted to much money in the past buying stuff I thought I needed only to find out I needed something else or something more to get done what I wanted to do. IF you can understand what I am trying to learn, please help me out.
 
I would like four XLRs on an interface with at least two powered but I do have sone XLR to TRS adapters... I do not want to consider an RCA as an input for what I plan to do. A couple of RCAs for possible unforseen future application might be a goodidea but I don't want to have to depend on them for what I want to do now.
 
To recap, my main questions are about when RTL is a concern, How much f a concern are the mic pres (especially regarding the interfaces listed by Jim and the Octa), should I be thinking of a KB midi controller to run through an interface using the interface drivers and RTL or should I be thinking about a KB controller to connect via USB and its ability to control X1 using ACT... and will the USB connection have a fast an RTL as it would if it were running through one of these interfaces?
 
Thanks for the help. 
 
Julien

Thanks Jim.

Julien