I managed to find a solution of your problem by simply using a disk-like structuring element instead of a square/rectangular one for the closing operation. A comparison of both approaches' results is shown below.
I am not entirely sure why the disk-like structuring element outperforms the square/rectangular one in this particular case. I will leave it to you to decide whether you will dig into the theory behind it or not. Of course any pointers in the comment section are more than welcome.
Here is the complete code:
import cv2 as cv
img = cv.imread("WhGXN.png")
disk = cv.getStructuringElement(cv.MORPH_ELLIPSE,(13,13))
square = cv.getStructuringElement(cv.MORPH_RECT, (13,13))
disk_img = cv.morphologyEx(img, cv.MORPH_CLOSE, disk)
square_img = cv.morphologyEx(img, cv.MORPH_CLOSE, square)