For a portfolio with normally distributed geometric (log) returns, the worst-case geometric return at a 95% confidence level is calculated using the Z-score for 95% (1.645): $R_{G}^{*} = \mu - Z \times \sigma = 0.12 - 1.645 \times 0.32 = 0.12 - 0.5264 = -0.4064$

The worst-case portfolio value is: W^{*} = W_0 \times e^{R_{G}^{*}} = \`85 \text{ million} \times e^{-0.4064} = \85` \text{ million} \times 0.6660 = \`56.61` \text{ million}$

The lognormal VaR is the difference between the initial portfolio value and the worst-case portfolio value: VaR = W_0 - W^{*} = \`85 \text{ million} - \56.61` \text{ million} = \`28.39` \text{ million}$

Thus, Option D is the correct answer.