The mechanics of a mismatch are the same, but the quantities compared differ — so on one chart MACD and RSI produce different sets of divergences. Neither is "more correct": they have different units, different thresholds, and MACD carries an extra rule.
What is actually compared
RSI — the oscillator itself, usually period 14. Values run 0 to 100, and divergence is measured directly in its points. A very small difference between peaks is noise rather than divergence.
MACD — not the line and not the signal, but the histogram: the difference between the MACD line (12/26) and its signal (9). The histogram is measured in price units, so a meaningful threshold for it is not absolute but tied to the instrument's volatility.
Hence the first consequence: RSI has its own scale, the same on every instrument, while MACD is measured in prices — so its thresholds cannot be compared across instruments directly.
The MACD zero rule
The histogram has a condition RSI does not: between the two peaks being compared, the histogram must cross zero.
The point is to keep the peaks in different impulses. The histogram is noisy, and one impulse is easily split into two humps; without the zero rule those humps would be compared against each other and produce a divergence inside a single move. This is Elder's rule, and it is switched on here, including for hidden divergences.
RSI has no analogue: it does not oscillate around zero but lives in a 0–100 band, and splitting it into impulses at some level would be arbitrary.
Why the sets do not match
The price extremes are the same for both — one detector finds them. Everything else differs: the indicator's value at the pivot, the divergence threshold, the presence of the zero rule.
So it is perfectly normal for RSI to see a mismatch at a pair of peaks where MACD sees none, or the reverse. That is not a contradiction: the two indicators measure different things.
Which to choose
The question is put wrongly, as with "FVG or order block". They are not alternatives but two independent measures of the same stretch. Both are worth looking at, and it is worth noticing separately when they disagree — that disagreement is more informative than either mark on its own.
What neither of them promises
Neither by MACD nor by RSI is a divergence an entry signal, and neither is used as one here. It describes what already happened: the current move's impulse is weaker than the previous one by that indicator's measure.