While derivative pricing models are better modeling reality as academia invests more into the subject, none sufficiently do.
If, for example, one assumes that stock returns are lognormal for the purposes of pricing options like Black Scholes does, the only true dependent variable becomes log-standard deviation otherwise known as "volatility", producing the infamous "volatility smile" which disappears in the cases of models with more factors accounting for other mathematical moments such as mean, skew, and kurtosis, etc. Still, these more advanced models are flawed, and suffer the same extreme time mispricing as Black Scholes. In other words, one can model anything however one wants, but the worse the model, the stranger the results since volatility for a given expiration should be constant across all strikes and is with better models.
In the case of pricing dividends, these can be adjusted for the many complexities of taxation, but the model becomes ever more complex and extremely computationally expensive for each eventuality. Furthermore, with more complexity in any model, the likelihood of discovering a closed form in the short run is less.
For equities in a low interest rate, not high dividend yield, not low volatility, low dividend tax environment, the standard swap pricing models will not provide results much different from one where a single low tax rate on dividends is assumed.
If one is pricing a swap on equity outside of the bounds above, the dividend tax rate could have more of an effect, but for computational efficiency, applying a single assumed dividend tax rate would be optimal with D*(1-x), instead of D in a formula, where D is the dividend paid and x is the tax rate.
In short, a closed form model is only as good as its assumptions, so if anomalies appear between the actual prices of swaps in the market and a swap model then that model is less correct than the one with smaller anomalies of the same type. In other words, if pricing equity swaps without a dividend tax rate factored more closely matches the actual prices than pricing with dividend taxes factored then it could be assumed that pricing without a dividend tax factored is superior. This all depends upon the data, and there doesn't seem to be much in academia to assist with a conclusion.
If equity swaps do truly provide a tax advantage and both parties to a swap transaction are aware of this fact then it seems unlikely swap sellers wouldn't demand some of the tax advantage back in the form of a higher price. A model is no defense since volatility curves persist despite what Black Scholes says they should be.