I have read some articles mentioning that a full dihedral reduces the lift produced by a wing. So what about a tip (or cranked) dihedral that has the angled part starting from the mid-wing rather than the root of the wing?
The ratio of vertical lift to total lift in a bank angle is cosine bank angle. Cosine of 5 degrees is 0.996. Dihedral "banks" each wing very slightly from vertical. Typical dihedral angles affect lift very little.
Working this problem a bit further, we find that the cosine function is non-linear as angle increases. This means that half a wing at a steeper angle with half a zero angle actually has less vertical lift than a full wing dihedral of the same fuselage to wing tip angle.
A "cranked dihedral" would also be more complicated to build and not as strong. The design is sometimes seen in older model gliders. Some full scale aircraft may have winglets, but are plenty stable without extra dihedralling at the wingtips.
The main performance penalty incurred by dihedral is arguably not the slight loss of "verticality" in the lift vector generated by each wing, but rather the fact that if any accidental sideslip is present, the angle-of-attack of the "upwind" wing will be increased and the angle-of-attack of the "downwind" wing will be decreased. (That's how dihedral "works" to create roll stability, after all.) Even disregarding the additional drag incurred by whatever aileron input is needed to cancel this roll torque, due to the shape of the curves of the lift and drag coefficients (versus angle-of-attack), flying around with the two wings at different angles-of-attack is generally less efficient than flying around with the two wings at the same angle-of-attack.
Making the wing flat in the middle and reserving the dihedral for the tips-- where it will have the most benefit in terms of maximizing the stabilizing roll torque generated by a sideslip-- minimizes this drag penalty during accidental (and perhaps transient) sideslips, for any given desired degree of overall "effective dihedral", i.e. any desired amount of roll torque generated by a given sideslip angle.