I believe I understand the following (from the accepted answer to the Quantitative Finance question called "volatility of a mid curve option"):
A swaption in which the underlying swap starts at a date materially after the expiration date is called a midcurve swaption. The implied volatilities of these can not be obtained from the regular swaption surface. Market makers calculate implied volatilities for midcurves in a number of ways. One popular method is to compute the volatility of the forward swap using the volatilities of two spot starting swaps, and the correlation between them. For example , consider a midcurve option expiring in 1 year into a swap which starts 5 years later and ends 10 years later. The correct volatility can be computed from the 1yrx5yr volatility, the 1yrx10yr volatility , and the correlation between 5yr and 10 yr swaps for the next year.
But, my question is, what strikes should be used to get 1Y5Y volatility and 1Y10Y volatility? Specifically, if we use SABR model, what is the strike to be input to handle the vol skew?