Hii guys I’m creating lighting variations on PORTA with a MEGA 2560 board.
I am using a zero-cross detection that triggers an interrupt, which calls the “void passage_a_0” function.
A command (continuous action on a button) modifies a lighting level setting from 1 to 25. This setting adjusts the phase delays corresponding to the lighting level.
I have an issue with the ISR interrupt. Below is the simplified code of the two related interrupts:

// Simplified zero-cross and ISR for continuous variation on bit 7 of PORTA
void passage_a_0 () {
  TCNT1 = 0;                           // Initialize TNCT1
  if (variation_demandee == 128) {      // We are on bit 7
    PORTA &= ~variation_demandee;       // Turn off bit 7
    OCR1A = retard_phase[niv_ecl[7]];   // Set the value of OCRIA
    bitSet(TIMSK1, OCIE1A);             // Set OCIE1A to 1 to allow the compare interrupt with OCIE1A
  }
  n0++;                                 // Count the number of passes in this interrupt
}

ISR(TIMER1_COMPA_vect) {
  PORTA |= 128;                         // Immediately act on the port with the concerned bit
  bitClear(TIMSK1, OCIE1A);             // Clear the OCIE1A bit to disable the compare interrupt with OCR1A
  n++;                                  // Count the number of passes in this interrupt
}

Problem:
With the `bitClear(TIMSK1, OCIE1A);` instruction in the ISR, which is necessary, the ISR is executed immediately as if OCR1A were set to 0.
However, if this instruction is removed, the variation works correctly, taking into account the OCR1A values for the phase delay. But depending on the value of OCR1A, the ISR is called several times before the next zero-cross.
Why is the ISR triggered immediately in the presence of ‘bitClear’?
Is it a code error or a defective board?