R^2 consistent with a rejected F-test

Consider the multiple linear regression model $Y_i = \beta_0 + \beta_1 x_{i1} + \beta_2 x_{i2} + \beta_3 x_{i3} + \varepsilon_i$, $i = 1, 2, \dots, 31$, where $\varepsilon_i$ are independent and identically distributed $N(0, 1)$ variables. The $F$-test for testing significance of regression rejects the null hypothesis $H_0: \beta_1 = \beta_2 = \beta_3 = 0$ against the alternative $H_A$: at least one $\beta_j$ is not equal to 0, $j = 1, 2, 3$, at 5% level of significance. It is given that $F_{0.05;\, 3, 27} = 2.96$, $F_{0.05;\, 3, 30} = 2.92$, $F_{0.025;\, 3, 27} = 4.01$, $F_{0.025;\, 3, 30} = 3.91$, where, for a random variable $F$ following central $F$-distribution with $(m, n)$ degrees of freedom, $P(F > F_{\alpha;\, m, n}) = \alpha$, $\alpha \in (0, 1)$. Then the value of $R^2$ cannot be equal to