ES-9 — fixed ~76 ms recording latency in Bitwig that doesn't change with buffer size (macOS)

[dmuirw]

I have an Apollo twin to ES-9 aggregate device with drift correction on on the ES-9 and Ive done the Aggregate device flipped around also ES-9 to Apollo etc. and nothing is working to fix the Latency. Ive also tried the WES-9 by itself and still no way to fix the latency. I've used Silent way and HW Clock plugin Bitwig.-Nothing

Problem in one line: Recorded audio from the ES-9 lands a fixed ~76 ms (0.076 s) behind Bitwig's grid, and the lag does not change when I change the buffer size. Bitwig's Recording Offset caps at 2048 samples (~43 ms), so it can't compensate the full amount.
System:

MacBook Air, [16 gb. M1], macOS [Seq 15.7.5]
Bitwig Studio [6.1]
Expert Sleepers ES-9 (class-compliant USB on macOS), firmware [latest]
UA Apollo Twin [apollo Twin/ X ], run together with the ES-9 as a macOS Aggregate Device
Clock to modular via Bitwig HW Clock Out → ES-9 → ALM Pamela's PRO Workout (24 PPQN) + WMD Metron. The clock works perfectly — this is only about recorded-audio alignment.

How I measure it:

Loopback: ES-9 output → ES-9 input with a patch cable, play a sharp click on beat 1, record it back, measure where it lands. Result: ~0.069–0.076 s late.
Real-world: a modular voice clocked by Bitwig, recorded against a Bitwig kick on the grid. The modular lands 0.076 s behind, measured transient-to-transient.

The key anomaly — it's fixed across buffer sizes:

512 samples → ~0.069–0.076 s
256 samples → ~0.076 s
128 samples → ~0.076 s (identical; a 4× buffer reduction changed nothing)

Real I/O latency should scale with the buffer. Mine doesn't move at all.
Already tested and ruled out:

Aggregate device order (ES-9 first vs Apollo first): no change
ES-9 alone, no aggregate: same ~0.069 s
Drift correction off: no change
Both devices confirmed at 48 kHz: no change
"Audio Recording Latency Compensation" is ON
Record track, click track, and master bus all empty (no plugins / no PDC): no change
Clock is via the HW Clock Out device (not HW CV Out, not a Grid CV Out patch)
Recording onto a plain audio track (not a CV In module)
Recording Offset maxes at 2048 samples (~43 ms) — can't reach the ~76 ms

Constraint: I need to run the ES-9 + Apollo Twin as an aggregate, because the Apollo handles my monitor outputs and vocal/instrument tracking. So ES-9-only isn't a usable permanent setup.
Questions:

Why is the ES-9's record latency a fixed ~76 ms that ignores the buffer size? Is the macOS generic USB driver under-reporting the ES-9's true latency to Bitwig, so it can't auto-compensate?
Is there a fix — a config/firmware/CoreAudio/aggregate setting — that either reduces the underlying latency or makes it report correctly so recordings land on the grid automatically?
If not, is there a recommended routing or workflow for this exact ES-9 + aggregate + macOS situation?

[BobDog]

if the ES-9 when used alone is giving you 69ms then I would start looking there first, it seems a huge amount.

[dmuirw]

Yes, Ive tried everything and the same result. Ive worked with Claud and Gemini on ever option and still nothing. Its so bizarre bc my friend has the same set up and it works.