Improving cross-border capacity for near real-time balancing

European markets continue to integrate with two ongoing projects for common markets of manually activated balancing reserve. In the design of these balancing markets, the method for computing cross-border capacity has several limitations: it uses a network representation forecasted too early, uses a methodology that does not represent well meshed grid physics and does not properly account for remedial actions put in place by the grid operator to mitigate congestion. In this paper, we provide an improved capacity computation process for manually activated balancing reserve using a flow-based methodology and a network representation forecast one hour before real-time. We assess the impact of the current and the proposed capacity computation processes on market prices and network security configurations on an updated IEEE-96 network. Results show an increase in social welfare with the proposed methodology.