Plot the phases for the three voltages and you'll find that they're out of phase at the crossover points:
The problem here probably is the phase of the midrange driver. Focus on that. You should also take into account the actual phase of each of the three drivers (which are just 8 Ohm resistive loads in your simulation).
Here is the result assuming a simple phase reversal (180-deg) in the midrange driver:
And here you have the total power delivered to the 3 loads, as suggested by The Photon, where you can see the approx -6dB power down at the crossover frequencies:
Also note that you may be looking after two things that are quite different from each other:
If you want to predict the frequency response that you would measure with a microphone, you need to factor in the response of the drivers (you just can't add them together), their physical positioning, etc.
If you want to predict the frequency response of the overall load seen by the amplifier's output, then you need to replace the 8 Ohm loads by the equivalent impedance of each driver.
1 and 2 can be very different things to simulate and measure, and will be highly dependent on the characteristics of each driver.