Specifying Redis port seems to have no effect

I was playing around with the GA charts, and configured my deploy to use an external redis. So far so good, with my config looking like:

global:
...
  redis:
    host: "external.redis.compose.com"
    port: 38474
    username: admin
    password:
      secret: gitlab-redis-secret
      key: redis-password
...

So I ran a dry run of the deploy, and happened to notice that pretty much every place the redis hostname is mentioned, the port is the default, like so:

workhorse-config.toml.erb: |
    [redis]
    URL = "tcp://external.redis.compose.com:6379"
    Password = "<%= File.read("/etc/gitlab/redis/password") %>"

or:

resque.yml.erb: |
    production:
      # Redis (single instance)
      url: redis://:<%= File.read("/etc/gitlab/redis/password") %>@external.redis.compose.com:6379

Then when I actually deploy it, the unicorn deployment will fail to start / schedule pods. So I checked the logs, and in the workhorse logs you will see piles of the following error (notice its trying to connect to the default port, not what I specified):

time="2018-08-23T17:02:05Z" level=info msg="redis: dialing" address="external.redis.compose.com.com:6379" network=tcp
time="2018-08-23T17:02:20Z" level=error msg="unknown error" error="keywatcher: dial tcp 10.10.10.10:6379: getsockopt: connection timed out"

Am I missing something here? Maybe the doc is out of date and the port cannot be specified in the global section?

For what its worth, I also tried setting the port as a deploy param like so (which had no effect):

helm upgrade --install --values=test-values.yaml gitlab . \
--set global.redis.port=38474 \
--timeout 600 \
--debug