Accepted answer
The commonly quoted figures for lean-mass loss are mostly measurement artefacts, and the artefact is well understood: fat-free mass as measured includes water and glycogen, both of which fall early and neither of which is contractile tissue.
Protein target, worked: at 88 kg, a target of 1.6 g/kg is 88 × 1.6 = 141 g per day. Spread across three eating occasions that is roughly 47 g each, and the leucine threshold for a maximal muscle protein synthetic response is met at around 2.5 to 3 g of leucine, which corresponds to roughly 30 to 40 g of a high-quality protein. So three meals at 40 g plus one 25 g snack gets you to 145 g and clears the per-meal threshold each time. On 900 kcal that leaves about 340 kcal for everything else, which is the actual constraint.
Hydration state moves a DEXA lean-mass figure directly, because the algorithm assigns water to the lean compartment. Scanning fasted, at the same time of day, before training and without a recent high-carbohydrate day is the difference between a comparable sequence and a noisy one. Bioelectrical impedance is far more sensitive to hydration again, which is why its trend is unusable at this timescale.
The body-composition substudies in the major programmes consistently report that the proportion of weight lost as fat mass is approximately three quarters or better, with the lean-mass fraction falling within the range seen in dietary weight loss of comparable magnitude[1].
Worth stating that a DEXA sequence is only as good as its protocol consistency, and most people’s sequences are not consistent enough to support the conclusions drawn from them.
Train, eat the protein, measure something functional, and give the trend three months before you interpret it.