Do the MONIAC's £400 build cost and ~2% economic-accuracy figures rest on a primary source?

The note claim-phillips-moniac-hydraulic-economic-computer-1949 carries two quantitative figures that, per the vault's sourcing floor, require Tier 1–2:

• the ~£400 cost to build the 1949 prototype, and
• the claim that it modeled UK national income flows to within roughly 2% of real data.

Both currently rest only on the Tier-4 Wikipedia article Phillips Machine. Quantitative claims are exactly what secondary retellings garble, so these are flagged [unverified-quant — needs primary].

What would answer it:

• Phillips's own account — A.W.H. Phillips, "Mechanical Models in Economic Dynamics," Economica 17 (1950) — for the machine's design and any accuracy claim in his words.
• The Science Museum Group object record (object no. 1995-210) or its curatorial notes for a costed, dated institutional figure.
• Nicholas Barr's or Walter Newlyn's historical accounts of the LSE machine, which discuss its construction and calibration.

Until one of these is read, the figures stay flagged and the note stays seedling.