There is obviously no non-teleological answer for this, but it can be much much more straight forward than metabolism.
Think of plants as basically solar panels. We can think of the 'grass' phenotype as competing through r-type selection. Lots and lots of individuals, lots of surface area, short life cycles. Short bursty growth, convenient loss resilience, easy mobility.
'Trees' are the also optimizing for surface area (among other things), but are relying on k-type selection. Longer life-cycles, lower probability of successful reproduction, specialized pollinators, and very significantly, vertically stratified growth habit. Being an individual in a k-type species is like a very significant advantage for that individual, since its unlikely that the vast majority of k-type seeds will successfully reproduce.
Being k-type may not be more competitive for the species overall, but it may result for a selection bias at the individual member resolution, and so may be unlikely to be selected against.
There are many many advantages to lignification when your whole thing is basically being solar panels on a pole. It would seem extremely unlikely this has anything to do with metabolism other than 'doing more of it'. We already know why monocots don't make trees more significant than palms (read: vascular bundles and eusteles). My money goes to species level versus individual level selection in k versus r.