This is probably getting too deep and should move to the cap thread. I believe what you're saying is correct, but at audio frequencies the esl is canceled by the capacitance. In terms of a black box model, it doesn't exist at any one frequency, even if you draw it as part of the model. I believe it's only an explanation for the overall impedance curve, and shows up as a real value beyond resonance. That's way too high to worry about in a crossover. The thing that bothers me is the definition of esl. The data sheet says they measure it at resonance, but that doesn't make sense. At resonance, the phase shift should go to zero. In fact, that's how you find it with any accuracy. If the phase shift is zero, you don't have any reactance at all, just a pure resistance. So where's the esl at that point?
No doubt the poor OP lost interest in this way back, but I'll say most of the crossovers I've built had parallel caps for two reasons. First, I never had exact values available, so parallel combinations were needed to get what I wanted. Second, big values are expensive or unavailable, so I often paralleled inexpensive and surplus caps to get up to the larger value needed. FWIW, I've found parallel combinations to be indistinguishable from single caps of the same value and type on any measurement system. IOW, both value and losses remain the same.
