It may not be the comprehensive answer you are looking for, but it's very common for larger companies to offsource the majority of their infrastructure gruntwork, including the handling and building of plain IP blocklists in response to perceived attacks.
If you look for IP reputation, you'll find that there are lots of companies offering such a service, and they are very attractive for larger companies, as they can show that they spend the money on these subscriptions for an unavoidable purpose.
One well known, easy to find company states:
An IP address earns a negative reputation when (OURCOMPANY) detects suspicious activity, such as spam or viruses originating from that address.
Now it is easy to conceive there does not need to be anyone involved in any malicious activities. It's just their machine may be compromised and actually try to scan open relays or probe for exploits on other hosts without even their owner's realising it.
So yes, you can be sure there's a noticeable proportion of traffic originating from Tor exit points that can be considered malicious. Considering this, it cannot be a problem that any tor user can avoid by tweaking settings.
From this point on, it's not a technical, but a political problem - how much risk can we ask from companies to take when accepting traffic from the IP addresses we share with potential attackers?
And then it's getting very mushy - some people are offended by open relay checks against their not relaying SMTP server, and consider this an attack - whereas in reality, it might be 1kB traffic wasted - in a day with potentially gigabytes of traffic on an XXMbit/s link. Who knows where to draw the line?
I'm not even sure we can convince any of those larger companies to make exceptions for TOR endpoints, as they have legitimate reasons (avoiding country based blocklists), that are against the policies, and wishes(!) of those very companies.