How to add the two output of a PWM IC

It might be possible that the PWM input is only a controlling input for the fan which means that some hidden circuits inside the fan do the "real" controlling of the fan and maybe also vary the voltage together with the duty cycle. I've seen something like this before.

The TL494 can be used as a PWM generator. It's not clear what frequency you need, but the part is not fast. It might generate usable PWM up to about 30kHz. After that switch overlap can become a problem. But, following your questions:

1. It's probably not a good idea to drive the feedback pin directly with a pot wiper. Better would be to choose one of the error amps, make it unity gain, and then feed the pot wiper into the non-inverting input. That takes the loading off of the pot, and lets the amp do what it was meant to do.

2. The error amp you don't use needs to be biased on the input so that the output is low. Tie the inverting input to Vref through a 1KOhm and tie the non-inverting input to ground through a 1KOhm. DTC sets the dead time on the PWM waveform, with low being maximum PWM (minimal dead time), dead time increasing as DTC voltage rises. For maximal PWM, keep DTC below about 100mV.

3. Vref can be used to supply voltage, but has to be done carefully. Vref can't source more than 5 or 10 mA, so any more current than that and a buffer BJT (pnp) would be needed to supplement. Unless the extra load is very constant, more filtering would probably be needed. Vref is used inside the chip, so any noise added to it could mess up operation.

4. Normally to get a single output, one would connect E1 and E2 to ground and tie C1 and C2 together with a single pull-up resistor as a wire-or.

The TL494 is a bit unusual in that PWM duty cycle is inversely proportional to the feedback voltage, when feedback voltage is low PWM is maximal, as feedback voltage increases PWM duty cycle is reduced. If you haven't read "Designing Switching Voltage Regulators with the TL494" you should.