^ Have you had a chance to read this article:
This is using the Neumann head (Tyll's setup), but I'm pretty sure that Audeze uses a similar head. While their setups are not identical (I would venture to say that Tyll's is likely more accurate/precise), the results show that beyond 4kHz there can be some dramatic swings in frequency response with the EXACT same pair of headphones due to very slight positional changes on the head. Tyll goes a few steps further by measuring 5 positions on the head and then averages them to help "smooth" out the data. Audeze DOES NOT do this. So their swings would be far greater (also documented in the above article).
So my guess (and money) are the differences you're seeing is very minute changes on positioning of the LCD-3s on the dummy head as they are all above the 4kHz range.