It is extremely difficult to provide any sort of flat "responsive design support" fee, percentage, or even algorithm, because in practice the devil is very much in the details.
The problem is (and what you absolutely must define in writing), what constitutes a "supported device"? By the gods, do not permit "mobile devices" and leave it at that. My first foray into that world was a trip to the school of hard knocks, and I'm endlessly grateful I wasn't on an hourly, contract, or time pressure.
The thing to know is you must treat each device and operating system version very much like you would an additional browser. As you are use to supporting multiple browsers, I'm sure you are familiar how much work can go into something as simple as "also, I want this to support Internet Explorer 7"; this can be easy, or it can be an amazing head ache depending on what technologies you are using.
What about Windows RT and Surface? How about older Androids, or the first/2nd generation iPad? iPod touch? What tablet sizes and models must be supported? Will you need to test Firefox and Dolphin and Opera for mobile devices?
Meanwhile, W3C validators and Adobe Browser Lab tools all reported no expected problems.
Bottom Line Advice
A multiple of 1.5-3X, depending on amount and age of devices supported, is not unreasonable. The only way to get better is to do it and get your hands on as many actual devices as you can manage and test, test, test. Its the new version of 1990s browser wars where the rules are made up and the points don't matter, you'll inevitably have to get dirty and spend whole days trying to fix one stupid thing that should work and works in 20 other platform tests but not on this one particular thing....
So define your terms, limit the officially supported devices (and specify models!), browsers, and configuration, and be prepared to feel like a bug-hunting noob. You an do it, just be prepared to spend significant amounts of extra time, avoid taking more than one job at a time to start with, and give yourself some intelligent leeway to avoid looking bad.
I'm personally a big fan of picking the most important configuration and delivering that, then locking down the other configurations on a staggered timetable. If it's all or nothing without incremental delivery you are asking for an unhappy client.
Most of all: good luck!