What you have there is a classical Switch Loop
You may be familiar with switch loops, where the power comes into the lamp (not the switch) and /2 cable is used to bring only supply hot and switched-hot down to the switch. You have a slight variation for "2-device, independent control". It involves supply hot, switched-hot 1, and switched-hot 2. This is most often seen on fan/lights.
The classic problem with switch loops is no neutral. Supply hot is present, but supply neutral is not, so there's no way to provision a device that needs 24x7 power, like a receptacle or simply a powered switch. You're not having trouble finding it; it's simply not there.
So Code was changed years ago to require a neutral be pulled on switch loops. Now 1-device loops use /3 cable (Hot Neutral Switched-hot) and 2-device switch loops use either /4 cable (H N SwH1 SwH2) or dual /3's (H N SwH1) (H N SwH2).
Currents must be equal
You can't use two /2 cables as a substitute for /4 because of a fundamental rule: Currents must be equal in every cable or conduit. If you have supply hot coming down cable 1 and returning via cable 2, that will create eddy current induction heating, wire vibration and fatigue, and all sorts of other fire-starting badness.
This "currents must be equal" rule requires the dual /3's where each /3 cable serves only its own switch. This assures that hot (for switch 1) equals neutral (for Sw1) plus switched-hot 1. Currents are equal since H1=N1+SH1.
So the options I see are:
- Gang both banks of lights off 1 dimmer (cost -$64: send back a dimmer; buy blanking plate)
- Scrap out the /3 and replace with /4 (cost /4 cable + fishing labor)
- Lay another /3 parallel to the first (cost /3 cable + fishing labor)
The /4 is most elegant since it lets you use correct wire colors natively: White neutral, Black supply, Red switched1 and Blue switched2.
Or the dual [or single] /3 option:
Most consumer-tier dimming isn't very good. It's a very crippled technology by nature, because the dimmer must be cheap, it must use existing wiring, and it must interoperate with an impossible variety of light bulbs. And so they take the AC supply sinewave and they horribly chop it up - why? Because it's easy with triacs. This creates poor power factor and lots of line noise on the supply side, and "hum" on the load side. Lots of sites talk about whether "leading edge" or "trailing edge" are better, but both are mutilated sinewaves. This isn't how you should control lighting.
And believe me, LEDs have an inherent, built-in ability to dim to absolutely any level -- limited only by the driver circuit and the dimmer knob's ability to talk to the driver. The mutilated sinewave methods are controlled by the need to be backward compatible with incandescent screw-in bulbs, and that makes them far too coarse and hanky-janky for any kind of precise control.
If we walk away from the screw-in bulb, we don't need to be backward compatible. And decent LED emitters are good for 30 years+ and are the least likely part to fail, so who needs sockets? Now we're free to design dimming architectures de novo, that let us draw out the full ability of LEDs. Here are some examples.
0-10V dimming -- first, industry is way out ahead of us. They've had dimmable fluorescent ballasts for an age. Not CFL screwballs! Real fluorescent fixtures. I just saw some at a hospital that were absolutely unbelievable, they were just perfect. Anyway, the gist is, a 0-10V dimmer on the wall sends a (you guessed it) 0-10V signal to each ballast (on the purple/gray wires). This is Class II wiring so it can just be thermostat cable tossed in the walls.
From the opposite direction (cheap low voltage lighting), we have PWM dimming specifically for LEDs. In this system, the dimmer knob sends a pulsed DC signal at several hundred Hz - far too fast to see. The duty cycle decides the brightness. With the right dimmer knob, control can be extremely fine. This signal goes to an amplifier which drives the LEDs proper, which run on low-voltage DC. If power is sourced at the lamp (as in your case), the AC->DC power supply is up there, as is the amplifier. The cable to the dimmer switch is simple Class II thermostat wiring.
Both of these dimming schemes can come very close to zero brightness. They can also strike (start up) at very low settings - you don't need to crank them up to 50% to get them to light initially.
Of course there are other dimming schemes, but those are two. PWM isn't even particularly expensive.