Suppose $f:[0,1]\rar[0,1]$ is measurable and $X_1,X_2,\ldots$ an i.i.d. sequence in $[0,1]$ with distribution $\mu$. Define $$ N\colon=\inf\{2n:\,X_{2n-1}\leq f(X_{2n})\} $$ then $$ \P(X_N>t) =\int_t^\infty\mu((0,f(x)])\,\mu(dx)/ \int_0^\infty\mu((0,f(x)])\,\mu(dx)~. $$ If $\mu=\l$, then $X_N$ has density $f/\int f\,d\l$.
Since $[N=2n]=[X_{2n-1}\leq X_{2n}]\cap\bigcap_{j=1}^{n-1}[X_{2j-1}>X_{2j}]$, we get by independence: \begin{eqnarray*} \P(X_N>t) &=&\sum_{n=1}^\infty\P(X_{2n}>t,X_{2n-1}\leq f(X_{2n})) \prod_{j=1}^{n-1}\P(X_{2j-1}>f(X_{2j}))\\ &=&\sum_{n=1}^\infty\P(X_{2n}>t,X_{2n-1}\leq f(X_{2n})) \P(X_{1}>f(X_{2}))^{n-1}\\ &=&\P(X_{2}>t,X_{1}\leq f(X_{2}))/ (1-\P(X_{1}>f(X_{2}))) \end{eqnarray*} Thus the assertion follows from $$ \P(X_{2}>t,X_{1}\leq f(X_{2}))= \int_t^\infty\mu((0,f(x)])\,\mu(dx)~. $$