Bar plot has bad additive color algebra when opacity<1

I need to plot a few histograms and for aethetical reasons i prefer to make them semi-transparent and in overlay mode. I have a problem with how plotly handles color additions. I am aware that there is an additive color scheme and subtractive color scheme. I am not sure what scheme in used when plotly combines semi-transparent colors, but the result is absolutely unsatisfactory.
What i expected:
Given 2 original colors "#FD3216", "#00FE35". Mix them in 3'd color like in this site, then display it with given opacity. I expect it to be "#7e9825", but with opacity. Same for dark and light modes.
What i got:
In dark - "#37BD2B", in light "#4DD341". In dark mode the color is almost indestinguishable from one of parent colors, for almost any color sequence.

Here is the code to reproduce the issue:

import torch
from plotly import graph_objects as go, colors

clr_seq = iter(colors.qualitative.Light24)
N_bins = 20
distributions = {'A': torch.normal(0, 1, [1000]),
                 'B': torch.normal(1, 2, [1000]),}
fig = go.Figure()
for k, v in distributions.items():
    fig.add_bar(
        **dict(zip('yx', torch.histogram(
            v, N_bins, range=[-5, 5]))),
        name=k, marker_opacity=0.7, marker_color=next(clr_seq))
fig.update_layout(barmode='overlay', template='plotly_white')
fig.show()

Commencez par la reproduction Python fournie en utilisant go.Figure.add_bar, marker_opacity, barmode='overlay' et template='plotly_white' ; comparez les couleurs de chevauchement rendues en mode clair et en mode sombre. Suivez le comportement de composition de l’opacité et des couleurs à partir de ce point d’entrée, et considérez l’issue comme terminée lorsque des barres semi-transparentes qui se chevauchent produisent systématiquement le mélange attendu documenté dans les deux modes.