True the load regulation is 12x worse than 0.5% on the secondary output, but it is not correct so say there is "no feedback" because the feedforward mode shares mutual coupling flux so the primary output regulation is adequate shared flux so that with 6% overall worst case error on the secondary output,which is including, line, load and ratio error tolerances when the current ranges or max power is not exceeded.
As for minimum current, this used to be manditory on old designs with inadequate phase margin. These designs have significantly improved for phase margin. Typcially the voltage would keep rising with no load. But the user wants 2-3 Amps and has not specified 0-3 Amps or tolerance, so we can only guess. Usually 1% to 5 % on 5V and 10% on 12V for most applications.
A mod to change the feedback point to 12V with a 5/12 voltage ratio is certainly doable if the OP is comfortable doing that.
In having qualified such designs from OEM's and worked directly with world class designers, from Hammond, Brown, Power One, Murata, Lambda, Shindengen ,I am fairly confident. but your mileage may vary.
I am familiar with epic failures from Packard Bell PSU, but not Freesat, I realize if the OP did the opposite, and loaded the 5V only with Pmax, we would expect the 12V to rise 4%, but in underdesigned or mismatched loads on multi-output feed forward converters, lack of mutual coupling flux can become catastrophic.
Meanwhile, Meanwell
has an excellent approach with hiccup, OCP, OVP, and OTP protection correction builtin.
Freesat market is often cost tradeoff with quality and HUMAX was known to use low quality caps.