I've swapped the LEDs around and found that the issue seems to be with the pins rather than the LEDs.
I would have thought the symptom would follow the leds rather then the pins. Either way, setting current level for different color leds is not as simple as it might seem at first.
First a formula for determining resistor value for a specific current value, say 20ma.
R (in ohms) = (source voltage - led forward voltage drop) / current in amps. The first thing to know is that different color leds have pretty significant different forward voltage drop. So a red led typically has a 1.5 Vf rating so R= (5-1.5) / .02 = 175 ohms. A blue led has, I think, a 3.5v Vf rating so R = (5-3.5) / .02 = 75 ohms. So you see if you want to drive a red and blue led at the same 20ma current they require quite different values for their series limiting resistor. Another issue is that different leds have different efficiency ratings so don't all put out the same amplitude of light at the same specific current. Yet another issues is that led light output is not very linear with current. That is many leds running at 10ma will not seem very much brighter if adjusted up to 20ma, while changing from say 1 ma to 10ma will see a much wider range of brightness change.
So you can see you have challenges ahead. When you send a analogWrite value of 255 (full on) to your three different leds, you will have drastic differences due to the use of a same size resistor for leds with different values of Vf, and you have no real idea I suspect of the current to light efficiency value for the three different color led.
So you might want to experiment a little. What I would do is calculate the correct resistor for each color using their specific Vf rating and setting the current level to 10ma. Then run your sketch and see if you don't get a wider range of brightness control. After that if you find all have good control but one color is hotter then the other two then just raise the resistance some for that one color. Using 10 turn 1k ohm trim pots can make the experiment go a lot quicker and after you get the results you want just measure the pots adjusted value and you then can obtain fixed resistors of the same or closest available value.
Sound like a plan?