It's all to do with the way motors work. The small motors you are using are constructed of a fixed circular magnet inside the metal case, inside that is a rotor with several coils on it and commutator. the commutator is basically a multi-way switch which takes power from the outside through brushes and routes it to one or more of the coils. The brushes are fixed and the commutator rotates along with the coils. As it turns the switch routes the current through different coils and keeps the electromagnetic field from the coils out of line with the fixed magnet. The N and S magnetic poles try to align with each other just like conventional bar magnets and in doing so pull the motor around on it's shaft, this is where the rotation comes from. As the rotor turns it also changes the current routing through the commutator so the current through the coil is turned off and the next coil is turned on, this pulls the rotor to the next N-S alignment and again switches the current to the next coil. So it constantly chases it's own tail and keeps rotating.
Now the clever part. As the motor turns it also works a bit like a generator. In fact if you measure the voltage across a disconnected motor and spin it you will see it generates a voltage across it's wires. This is similar to how power stations work but on a much smaller scale. The power it generates actually offsets some of the power it draws from the power supply, this is why the motor draws more current if you forcefully slow it down. If the motor is stopped completely (stalled) two things happen, the generator effect (look up Lenz's law) ceases and the current increases and also the commutator is very likely to have stopped in mid-switching position where more than one coil is connected, so twice the current tries to flow.
Starting from not turning, such as when you first apply the voltage, the motor needs extra current to overcome these two effects, this is why I suggest a capacitor might help because for a moment it would 'top-up' the USB current to give the motor a chance to get moving. When you added your resistor, what you did was restrict the power at the very time it was needed most. You might find that although the motor doesn't start by itself, if you spin it by hand you would overcome the initial starting power and it would keep turning by itself.
Given enough current, if you reduce the voltage the motor will run slower, unfortunately, a resistor does the opposite, it limits the current. Actually controlling the speed can be extremely difficult and requires a feedback system that monitors the speed and adjusts the voltage in real time. Probably the easiest way you can slow it's speed while still letting it start reliably and produce enough power is to use a variable voltage regulator, something like an LM317 regulator. You still need a voltage source and it has to be about 3V more than the motor needs (at full speed) but with only 4 components you could make a reasonably stable speed controller. Subject to there being enough current available from your power supply, the motor could then draw as much current as it wants as the voltage and hence speed is reduced.
Brian.