In a systematic review, Frankenfield and colleagues (2005) found the Mifflin-St Jeor equation estimated resting metabolic rate within 10% of measured values for roughly eight in ten non-obese adults and about seven in ten adults with obesity. That makes it the most reliable of the common equations, but it still misses for about one adult in five.
Activity is the bigger source of error. Levine and colleagues (1999) found that non-exercise activity differed by up to roughly 700 kcal a day between people eating the same surplus, so two people of the same size can have maintenance levels hundreds of calories apart.
The practical answer is a feedback loop: eat at the target for two to three weeks, compare your average weekly weight, and move calories up or down by 100 to 150 kcal if the trend is off.