We study spectrum auctions in India, where private operators acquire licenses through simultaneous multiple-round auctions (SMRA) while an all-India license is set-aside for the public operator. We quantify the government revenue loss under the set-aside and analyse the trade-off between revenue and public service coverage under counterfactual subsidized auction designs. Using a structural model of bidding accounting for both the stand-alone values and inter-license complementarities, we identify the model parameters from bidding behavior across licenses and rounds. We also introduce a railway connectivity-based index to capture complementarities for India. Counterfactual simulations show that allowing the public operator to bid without subsidy raises expected revenue by 7% and find an inverted U-shaped relationship between the public operators winning probability and circle level price changes, driven by heterogeneity in bidder valuations and complementarities. We also demonstrate that a location-based subsidy can yield higher revenues for a given level of coverage compared to a nationwide subsidy.