I have already told you what is wrong. You don't know how the caching works so please stop trying to be technical about it. I'm not working from pseudo knowledge at all.
The caches only receive requests sent to port 53 (DNS) and 80 (HTTP), no other traffic goes through them unless you explicitly send traffic to them by specifying a cache. Even this tends not to work too well if at all. So your claim of a 'default proxy' affecting your FTP download is just wrong. It's not really surprising that FTP doesn't download too well through a webcache.
Whether you specify or not, traceroutes cannot go through the ntl webcaches as they don't run over TCP. I have asked you repeatedly for info on latency / packet loss but again your obsession with cache 2 meant you ignored my question. They are web caches not all protocol proxies, misusing them as an all protocol cache as you are it's not surprising things aren't working 100%. They are dedicated webcache appliances supplied by Network Appliances.
There is no default cache, the IP address of the site you request is hashed and from this hash a decision is made on which of the caches to use, the entire IPv4 address space is split between the caches.
Frankly you're expressing your anger at the wrong people and overestimating your own knowledge here, it doesn't help that you're telling other people they don't know what they are talking about when infact you don't.
Just to repeat what's wrong, you have a bad upstream path to the ntl equipment, it's very unstable (as your two power readouts showing 61dBmV, the absolute max, and then 51.5dBmV have evidenced), and when a change in transmit power above a certain level is forced by a variation in this path it can cause a disconnect. Using a cache may alleviate this as it reduces the latency to servers dramatically meaning that there's more chance of retransmits compensating for the packet loss more. Could even be a load of noise on the upstream causing your modem to increase power to try and get over it too.
Ask the cretin you speak to at tech support (assuming it's one of the people who has so far failed to spot this extraordinarily obvious issue) to check the flap list on your modem, preferably looking for power adjustments, the column marked P-Adj. Their tools should be quite capable of doing this if he can be bothered to look at them. Also worth checking is your upstream SNR. If they can't see the problem (IE a very big number where working as intended modems have a very small one or zero) then they need the sack to be honest.
As you appear to have issues accepting that your technical 'knowledge' isn't all that here's a couple of links to save you the trouble of arguing the toss with what I've said:
Web Cache Control Protocol
Flap list info
Acceptable Modem power levels as per the sticky at the top of this board
Now stop trying to argue technically with tech support over which cache does what, how, when and why and just report a slow speeds / intermittent connection issue and ask them to check out the flap list for your modem. The fact it affects gaming as well as TCP based protocols, and affects FTP as well as HTTP makes it abundantly clear that caches aren't causing your particular issue.
I'm not saying that there isn't potentially a cache fault somewhere, but it's not the cause of all of your issues. Get the physical connection issues sorted first, then can move a bit further up the chain.
EDIT: For punctuation, clarity, and a small addition.