Applying ward method for images with the same labels

I ran the following code

dataset = datasets.fetch_development_fmri(n_subjects=1)
ward = Parcellations(method='ward', n_parcels=10,
                     standardize=True, memory='nilearn_cache', memory_level=1,
                     verbose=1)
ward.fit(dataset.func)
single = ward.labels_img_

single_plot = plotting.plot_roi(single, title="single",
                               display_mode='xz')

l = [single,single]
ward = Parcellations(method='ward', n_parcels=10,
                     standardize=False, memory='nilearn_cache', memory_level=1,
                     verbose=1)
ward.fit(l)
multi = ward.labels_img_
multi_plot = plotting.plot_roi(multi, title="multi",
                               display_mode='xz')

Since multi was computed using the same label image twice I would have expected the multi_plot to be exactly the same as single_plot. However the two plots are very different, what might be the reason for that?

the second one is fitted on the labels image, not on the same data as the first parcellation. why do you expect them to be the same?

Because the two files in multi contain exactly the same labels of single. If two images have exactly the same labels, let’s say

im1_labels = [1,1,2,2,3,3]
im2_labels = [1,1,2,2,3,3]

shouldn’t their combined clustering [im1_labels,im2_labels] be

im_combined_labels = [1,1,2,2,3,3]?

I don’t think that is necessarily the case. note the clustering is performed on a PCA decomposition after some preprocessing, not directly on the images themselves