Alarm not overriding template alerting rules

So I am trying to increase the thresholds for alerting for a specific label. So I dig through the documentation and come across that if a template and an alarm with the same name exists, the alarm will be preferred if both match. So from my understanding if I leave the original default config in peace and instead add a new config with an alarm that matches the alerts I want to modify with the correct labels this should override the alert limits.
However it currently looks like I also need to override the template and specifically exclude the labels to get it to work, otherwise it’ll just keep applying the normal rules.

Concrete example

/usr/lib/netdata/conf.d/health.d/web_log.conf:

 template: web_log_web_slow
       on: web_log.request_processing_time
    class: Latency
     type: Web Server
component: Web log
   lookup: average -1m unaligned of avg
    units: ms
    every: 10s
    green: 500
      red: 1000
     warn: ($web_log_1m_requests > 120) ? ($this > $green && $this > ($web_log_10m_response_time * 2) ) : ( 0 )
     crit: ($web_log_1m_requests > 120) ? ($this > $red   && $this > ($web_log_10m_response_time * 4) ) : ( 0 )
    delay: down 15m multiplier 1.5 max 1h
  summary: Web log processing time
     info: Average HTTP response time over the last 1 minute
  options: no-clear-notification
       to: webmaster

/etc/netdata/health.d/web_log_exceptions.conf:

    alarm: web_log_web_slow
       on: web_log.request_processing_time
    class: Latency
     type: Web Server
component: Web log
   lookup: average -5m unaligned of avg
    units: ms
    every: 30s
    green: 500
      red: 1000
     warn: ($web_log_1m_requests > 120) ? ($this > $green && $this > ($web_log_10m_response_time * 2) ) : ( 0 )
     crit: ($web_log_1m_requests > 120) ? ($this > $red   && $this > ($web_log_10m_response_time * 4) ) : ( 0 )
    delay: down 15m multiplier 1.5 max 1h
  summary: Web log processing time for CTAN mirror
     info: Average HTTP response time over the last 5 minutes of the CTAN mirror \
           (this alert is set to 5 minutes instead of 1 due to some individual \
           downloads taking a very long time!)
  options: no-clear-notification
       to: webmaster
chart labels: _collect_job=nginx_ctan-mirror

Now this does work, when I also include the web_log_web_slow template in the config and specifically exclude the _collect_job=nginx_ctan-mirror label.

Am I misunderstanding the documentation? Why do I need to override the template if I just want to modify individual instances?