Rasa test result inconsistent with manual test in rasa x

When adding FollowupAction in custom actions, rasa test has error predictions at places where different follow up actions should be triggered based on the slot values. However, the error is actually not happening in the manual test, is there a way to get rid of those false alarms?

class ActionCustomerCardIsDone(Action):

    def name(self) -> Text:
        return "action_customer_card_is_done"

    def run(self, dispatcher: CollectingDispatcher,
            tracker: Tracker,
            domain: Dict[Text, Any]) -> List[Dict[Text, Any]]:
        if tracker.get_slot("request_feature") == "circulation":
            return [SlotSet("customer_card_is_done", "True"),
                    FollowupAction("action_trigger_scan_qr_code_screen")]
        elif tracker.get_slot("request_feature") == "sale":
            if tracker.get_slot("customer_account_option") != "customer_card_already_have":
                return [SlotSet("customer_card_is_done", "True"),
                        FollowupAction("action_trigger_preference_screen")]
            else:
                return [SlotSet("customer_card_is_done", "True"),
                        FollowupAction("action_send_recommended_images")]
  - slot_was_set:
    - request_feature: sale
  - action: action_customer_card_is_done
  - slot_was_set:
    - customer_card_is_done: true
  - action: action_send_recommended_images  # predicted: action_trigger_scan_qr_code_screen

The example above is a single show case. Rasa test predicts next action by fully ignoring my conditions defined in custom actions which results in tons of double check in order to target the “real errors”. Could you please help? @koaning