Python: ski.transform.warp from scikit-image library does not work as expected

Question

I have recently started playing with some image processing and am relatively new to the field. I am at a lost as to why I get unexpected results when using the following code from the python scikit-image library to do a simple projective transform.

# TRANSFORM
imgGray = cv2.cvtColor(cv2.imread(imgPath), cv2.COLOR_BGR2GRAY)

dst = np.array([[1000, 100], 
                [1000, 200], 
                [800, 200], 
                [800, 100]])

src = np.array([[500, 0],
                [500, 200],
                [0, 200],
                [0, 0]])

tform3 = ski.transform.ProjectiveTransform()
tform3.estimate(src, dst)
tf_img_warp = ski.transform.warp(imgGray, tform3, output_shape=(200, 500))


## PLOT
plt.figure(num=None, figsize=(8, 6), dpi=80)
fig, ax = plt.subplots(1,2, figsize=(15, 10), dpi = 80)
ax[0].set_title(f'Original', fontsize = 15)
ax[0].imshow(imgGray)
ax[0].set_axis_off()
ax[1].set_title(f'Transformed', fontsize = 15)
ax[1].imshow(tf_img_warp)
ax[1].set_axis_off()


patch1 = Circle((dst[0][0],dst[0][1]),
                 10, facecolor = color)
patch2 = Circle((dst[1][0],dst[1][1]), 
                 10, facecolor = color)
patch3 = Circle((dst[2][0],dst[2][1]),
                 10, facecolor = color)
patch4 = Circle((dst[3][0],dst[3][1]), 
                 10, facecolor = color)
patch5 = Circle((src[0][0],src[0][1]), 10, 
                 facecolor = color)
patch6 = Circle((src[1][0],src[1][1]), 10, 
                 facecolor = color)
patch7 = Circle((src[2][0],src[2][1]), 10, 
                 facecolor = color)
patch8 = Circle((src[3][0],src[3][1]), 10, 
                 facecolor = color)

ax[0].add_patch(patch1)
ax[0].add_patch(patch2)
ax[0].add_patch(patch3)
ax[0].add_patch(patch4)
ax[1].add_patch(patch5)
ax[1].add_patch(patch6)
ax[1].add_patch(patch7)
ax[1].add_patch(patch8)

By reusing the code above, the results seem good at times e.g.

Other times, the results seem to have change the image in unexpected ways e.g. from the original image below, we can see some black tiles and some white patches on the top left corner of the selected area to be transformed.

However, the output from the transformation seems to alter this pattern.

As we can some of the black tiles are gone, together with the white patch on the top left corner. I have attempted multiple times to check for any silly mistakes on my end but I found none that is causing this unexpected result.

The output image has not been cropped, as can be seen from the number of rows and columns of tiles is the same as the selected area in the original image.

Can any image processing sifu out there help this newcomer understand what is going on? I am absolutely dumbfounded. Thanks!

EDIT p/s: I have also tried using opencv's perspective transform. But the results are the same as well.

M = cv2.getPerspectiveTransform(src, dst)
imgWarp = cv2.warpPerspective(img, M, (img.shape[1], img.shape[0]), flags=cv2.INTER_LINEAR)
imgWarp = Image.fromarray(imgWarp)