Because the example sentences will only use kanji that have been covered so far by the user, some words that are say normally written with two kanji might be written with only one, with the remainder of the word being written in hiragana. Thus words that would normally be recognized easily suddenly require an exercise in having to recognize that the hiragana probably map to a particular kanji or part of word and mental backtracking to what the word actually is.
Trying to go through 100 lessons while also reading the example sentences thus becomes a bit of a pain and I become tempted to skip the example sentences entirely for this reason.
I think a happy medium is to label unlearned kanji with furigana so that at least the word can be recognized by the kanji it’s actually written in.
The crux of the problem I think is that the way the sentences are written, excluding kanji that haven’t been covered yet, reads and feels really awkward and unnatural. I think for a Japanese learning site it’s okay to present sentences as they would normally be written.
For furigana to be viewed as a “crutch” in this case doesn’t make sense because the sentences should exclude furigana for kanji that have been covered by the user.