Saw this B&W answer in one of the FAQs at the support site:
We usually find that customers who alter crossover components are not fully satisfied with the results. They find that some aspects are improved, but others made worse. A classic case of this is when a polypropylene or other very low-loss type substitutes an electrolytic capacitor. We all know that polypropylene capacitors can sound inherently better, but the change in internal losses changes the response of the filter, which is designed assuming the losses of the electrolytic component. What usually happens when the low loss component is fitted is that the corners of the roll-off are sharpened, giving a peak in the combined response that can make the sound unpleasant in various ways depending on the crossover frequency. One way of getting round this is to wire a small resistor in series with the capacitor to approximate the original losses. I say approximate because the loss factor is a frequency dependent resistance. The actual value you need depends on the original capacitor loss factor and its capacitance value. The larger the value, the lower the resistance for a given loss factor. The formula for the equivalent resistance is:
R = d / 2