When I attempted docking I have annoying issue of RCS changing the ship speed. Once when I tried to rotate next to fuel station I got 10 m/s just by rotation.
I place RCS near COM when designing the ship - before adding spare booster/thrusters - just final module COM (well second to last if we count landing pod as stage).
The only difference between RCS thrusters and ordinary thrusters is the control scheme, the forces they apply work the same way:
Any time a thruster is fired, it applies linear momentum to the ship (which is the same as translation or changing the ship speed). The amount is independent of where the thruster is placed.
If the center of thrust does not align perfectly with the center of mass, it will also apply angular momentum (which is the same as rotation or applying torque) to the ship. The further away it is from the center of mass (CoM), the more angular momentum will be produced.
So, in the case of RCS, what you generally want is for it to be far away from the CoM, but balanced around it (placed symmetrically around the CoM). This makes both RCS modes work well:
If you use RCS to rotate, the thrusters on the opposite ends of the ship will fire in different directions, cancelling out the linear momentum, and adding up the angular momentum, resulting in rotation but unchanged speed.
If you use RCS to translate, the thrusters will fire in the same direction, adding up the linear momentum and cancelling out the angular momentum, resulting in no rotation, but changed speed.
Side note: If all you need is rotation, using reaction wheels is a reasonable alternative - reaction wheels never change the speed of the ship, only rotate it, and do not use any fuel, only electricity. They are more effective if you place them near the CoM, but will work anywhere, too. The trade-off is that they are bulkier and, for short missions, heavier (for similar power) than RCS, and cannot be used to translate (which is usually needed for docking).
Summary: Place RCS symmetrically around the CoM, but far away from it, for best results.
Do you place RCS in symmetric pairs/groups relative to COM?
To imagine the issue, let's take a very simple craft that has only one RCS block, somewhere off to the side. 
When you want to turn this craft any way using the RCS, there is nothing to offset the thrust. The craft will begin turning as requested, but it will also be accelerated in the direction opposite to the one RCS fires.
Also, KSP tends to fire all RCS at full thrust, so - say, you placed three thrusters in a 3x symmetry.
You want to pitch the craft's nose up. The one top thruster fires retrograde, the two bottom thrusters fire prograde. Net torque makes the craft pitch up. Net thrust - accelerates it forward; two thrusters forward, just one back.
Small asymmetries result in smaller errors, but they do result in errors nevertheless.
First, the closer it is to COM, the lower the torque it can exert; the more thrust it must put into rotating the ship.
And the more thrust it uses, the more all errors accumulate.
To find the optimal placement, you can use the RCS Build Aid mod, or just 'eyeball' it with help of the editor's symmetry.
Also, try to place the thrusters meant for rotation as far from CoM as you can without sacrificing symmetry. Or just disable RCS pitch/yaw/roll actuation controls and use only the reaction wheels for that.
One more thing: completely regardless of RCS, two crafts WILL drift relative to each other in orbit. After all, if they have exactly the same speed but are not in exactly the same spot, they are in two different orbits. If craft A is in a perfectly circular orbit, and craft B is moving at the same speed, but 50 meters farther from the planet, it's currently exceeding the speed proper for circular orbit proper for its altitude; it's at the periapsis of an elliptical orbit which will take it a good bit away from craft A - out towards the apoapsis and backwards as climbing, it loses orbital speed. This is completely regardless of RCS, and while unnoticeable in a short time period, in the 10 minutes it takes to move to the opposite side of Kerbin it can easily accumulate to 10m/s.
