cross-entropy

First network (what you want)

computed(c,d)  |targets(c,d) |correct?
---------------------------------------------
0.66 0.33  | 1  0     | yes
0.66 0.33  | 1  0     | yes
0.66 0.33  | 0  1     | no

avg cross-entropy loss: 
-(ln(0.66) + ln(0.66) + ln(0.3)) / 3 = 0.678335


Second network (what you said will yield a better network in terms of avg cross entropy)

computed(c,d)  |targets(c,d) |correct?
---------------------------------------------
1.0 0.0  | 1  0     | yes
1.0 0.0  | 1  0     | yes
1.0 0.0  | 0  1     | no

avg cross-entropy loss: 
-(ln(1) + ln(1) + ln(0.0001)) / 3 = 3.070113