Let $E/\mathbb{Q}$ be an elliptic curve. For $r\in\mathbb{Q}\cup\{\infty\}$ let us consider the modular symbols of $E$
$$\lambda(r):=2\pi i \int_{i\infty}^{r}f(z)dz\;\in\mathbb{C},$$
where $f$ is the modular form of $E$. By Manin-Drinfeld, we can write $$\lambda(r)=\left[ r \right]^{+} \Omega^{+} +\left[ r \right]^-\Omega^-$$
where $\Omega^\pm$ are the real and imaginary periods of $E$, and $\left[r\right]^\pm\in\mathbb{Q}$ are the so-called rational parts of the modular symbol.
$\textbf{I am confused by the following:}$
The modular symbols are essentially the modular parametrization $\varphi:X_0(N)\rightarrow\mathbb{C}/\Lambda_E$ applied to the cusps (before quotienting by $\Lambda_E$). So, for every rational number $r$, the denominators of $\left[r\right]^\pm$ should be killed by the number
$$D:=2\,\mathrm{lcm}\{ \text{orders of the images of the cusps of $X_0(N)$ in $\mathbb{C}/\Lambda_E$}\}$$
(the factor 2 is there just to account for the possible usage of $\Omega^+/2$ depending on whether we have 1 or 2 real components).
Now, let $E$ be the elliptic curve labeled $11.a3$ in the LMFDB: $$E:y^2+y=x^3-x^2$$
This curve has conductor 11, so it is semistable. Therefore the modular parametrization should map all the cusps of $X_0(11)$ to points of $E(\mathbb{Q})_\mathrm{tors}\subset E\cong \mathbb{C}/\Lambda_E$. Hence, since $\#E(\mathbb{Q})_\mathrm{tors}=5$, the integer $D$ above should be at most 10.
However, using SageMath we can see that $$\left[\frac{4}{7}\right]^+=\;-\frac{9}{25},$$ whose denominator is not killed by 10. Isn't this a contradiction?
Maybe I am misinterpreting or forgetting something. What is going wrong with my reasoning?
There is even a paper by Christian Wuthrich titled "Numerical modular symbols for elliptic curves", where in proposition 1 he proves that for a semistable $E$ we should have $$\left[r\right]^+\in \frac{c_\infty}{2\;\#E(\mathbb{Q})_\mathrm{tors}}\mathbb{Z},$$ where ${c_\infty}$ is the number of real components of $E$. So again, I cannot make sense of the denominator 25 above.