The percentages of machine translation modifications displayed inside "Pick a sentence" and "Preview & Publish" step do not match their actual values in some cases. More specifically:
a. When a block template exists inside the translated contents, the score of the unmodified MT validation is lower than the actual one. This happens because the "getProposedTranslation" method of the subSection model doesn't take into account block templates, but only sentences.
As demonstrated in the screencast, the unmodified MT percentage should be 100%, since the MT was not modified at all, but it's only 92%.
b. When different MT providers are used for the same subsection (i.e. paragraph), the MT modification percentage displayed inside the translated segment card is wrong, since the applied translation of the sentence is compared with the machine translation of the currently selected MT provider instead of the one actually used for this sentence.
As demonstrated in the screencast, the modified MT percentage is calculated correctly for the sentences that have used the same MT provider as the currently selected one, but wrongfully for the sentences that used a different MT provider and the whole subsection (paragraph).