It's actually not much to do with 12 hour displays but since I don't have a separate project for the display boards I made and the Github repo is linked to here, I'll log my experience here.

You may, like me, have purchased small sample bags of 50 or 100 LEDs for experimentation from AliExpress and similar marketplaces. Unfortunately I discovered that the quality can leave something to be desired. A bag of 3 mm green LEDs varied in brightness. This shows up in segment displays where the LEDs are in a line. Recently I discovered a bag of 5 mm blue LEDs also varied in brightness. It's hard to show in a phone photo but you can see there's one LED in the A segment of the hours digit that's noticeably dimmer than the others. The reason is not because the diodes were less efficient, but because they had stray resistance in parallel due to manufacturing defects. An ohmmeter showed finite resistance for bad specimens, but infinite for good specimens since the measurement voltage is below the LED threshold of about 3V. Worse still the finite resistance showed even when measured in reverse polarity. It's more noticeable for blue because these LEDs require less current so even a bit of parallel leakage matters.

My guess is that these substandard LEDs are bought by small traders and repacked for sale to experimenters. Mind you, if you are using them as single indicators, they'll probably work fine, perhaps with a bit more power draw. I have since moved on to purchasing bags of 1000 packed in the factory which should have better quality and consistency.