Most of the safety factors we use account for an uncertainty of about 30% in friction angle. A safety factor 3.0 in bearing capacity is about the same that a factor of safety 1.5 in retaining walls or slopes. This is so because the bearing capacity equation is higly nonlinear (Exp(pi*tan(fi)...). You reduce your friction angle by 30%, your bearing capacity falls by 60%. Hope it helps.

What is your opinion when it is purely cohesive soil ?

There are so many UNCERTAIN parameters in determining SBC, why emphasise Friction angle only?

Is not a good idea try to impose a safety factor in a general way. Try to control the risk only by one number is not the best choise I think....

Alejo's response may be based on that the bearing capacity factors (Nc, Ngamma, Nq) depend on the friction angle.  However, I think that most of the times, settlement controls the foundation design. So, the allowable bearing pressure which is determined by settlement analysis is lower than the bearing capacity (shear capacity) obtained by Terzaghi's or other equations.  Contrarily to the bearing capacity, settlement calculations do not use factors of safety since they are based on serviciability requirements. 

My answer, is for the originall cuestion.
Not to complement any ohters answers.

It is commonly assumed reasonable to use a higher factor of safety for bearing capacity as a way to limit settlement to acceptable levels.  Empirical evidence over many decades has shown that a factor of safety of 3 on bearing capacity usually limits settlement to acceptable levels.  This is rather simplistic and doesn't apply to all circumstances.

I agree that settlement almost always governs the final design capacity given.

if we evaluate the safety factor of the bearing capacity with a phi/c reduction method, the value recommend is 3?. 

Francisco, do you have references for that method?

a FEM model?