Unclear what proposition "No" relates to above. NOAA understands patterns of tides and strengths and limitations of LaPlace's Equations to an expert degree better than blanket denial suggests.
Continuing review of Lunisolar effects on data collection, tidal currents affect seasurface temperatures in complex ways. There is the internal sloshing. Shoals and other features promote mixing. Coasts, seamounts and islands have a sort of plunger effect. Given Chaos, there is both some coherent forcing, and highly chaotic outcomes. Some of this tidal churn will express as noise in sea-surface temperature data in the false guise of bulk forcing.
Surface Gravity Wave mixing is a huge source of seasurface temperature noise. A data buoy in heavy seas will read a different temperature than calm seas, in the same place and ENSO state. A data buoy will read a different temperature on a sunny day than cloudy day. As we dig down into the sensor dynamics, we'll get a truer picture.
Again, we see a pattern NOAA well knows, that ENSO identification and ENSO data is significantly uncertain. Lunisolar noise adds to the uncertainty, as well as having at best a weak theoretic (not well quantified) forcing effect. If ENSO Lunisolar Forcing was a strong effect, it would be long ago recognized and widely agreed. Training ML from inherently uncertain data is very imperfect.
Plate Tectonics is a clue. If Lunsolar tides where a source of harmonic order rather than noise, why are the continental plates so screwed up? It can't just be three-body chaos in vacuo, as a gross simplification, given Earth rotation as the first-order driver, not the incidental position of the three bodies. It really looks like seething multi-chaos, with a clean Lunsolar signal at most only superficially overlaid on the data.
Even if the 60Hz noise analogy is recast as a 59-61Hz dual beat cycle, it would not be forcing the music, only perhaps appear so, superposed.