Sharp GP2Y0A60SZLF module : unstable incomprehensible measure ?!

Dear
I'm built 7 Epoxy Board Breadboard (xbee, pro mini, rtc and 2 ir sharp 3v3 pololu) each board powered by 3 LS33600 batteries in parallel (3.6V, 51Ah, supported max pulse 250mA) with Schottky diode. On the board : with + The + the -.. With - nothing more.
I am a simple threshold crossing test.

wake_IR(IR1_pin_sleep); sleep_IR(IR2_pin_sleep); delay(10);
sensorValue1 = analogRead(IR1_pin_data);
sleep_IR(IR1_pin_sleep);

wake_IR(IR2_pin_sleep); delay(10);
sensorValue2 = analogRead(IR2_pin_data);
sleep_IR(IR2_pin_sleep);

if (sensorValue1>MAX)
detected_object_ir1++

if (sensorValue2>MAX)
detected_object_ir2++

On PC everything works perfectly but not on battery. Only 3 cards are walking good, others count over a factor 100 or more !! I train and do not understand, it's the same code, the same hardware, new batteries, the welds are average but on pc everything works, I do not understand and I can not find a line of research?
it should add resistance? capacitor? I do not know .
Philippe.

Can you change the battery pack of the good boards to the bad boards to see if it is a power problem?
Not sure about LS33600 but with rechargeable li-ion batteries voltage can vary from 4,3V (full) to 2,55V (empty)