Don't use NTP master. That is telling peers that your device is authoritative (even if it loses sync with its upstream server and is not). Apparently Juniper doesn't like that.
On the Adtran, configure the authoritative Cisco and a secondary peer or more if you choose with the command "ntp server <ip.add.re.ss>" and your Adtran will pull clock from it/them.
Point the Juniper at the Adtran's IP. It should just work. You should see the stratum of the source at the Adtran incremented by one. If the Adtran loses sync with the Cisco, it will no longer be authoritative and both its and the Juniper's clocks will free-run until sync is re-established.
You can also, in the Adtran, configure the Juniper as a peer with "ntp peer <ip.add.re.ss>". This will allow bidirectional synchronization which could be risky if the Juniper claims to be a lower stratum than what you see from the Cisco. This might happen if the Juniper is (mis)configured as NTP master.
An alternative, which I haven't tested in the lab, might be to use the ntp master command on the Adtran with a stratum higher than you get from the Cisco, something like "ntp master 8". This may allow the Cisco's clock to pass through to the Juniper if present. Untested. Obviously if the Adtran loses sync from the Cisco, the Juniper will ignore NTP from the Adtran as it is doing now.
