Not a daft question at all. There're probably a range of factors.
From Ben Goldacre's paper[0] pharma industry sponsors are more likely to be compliant than non-industry sponsors, and sponsors of large trials more likely to be compliant than sponsors of small trials. This may suggest that it's partly a matter of human resource to actually compile the data and report it onto the public websites in question. Here, I'd also consider Hanlon's razor - busy people with other priorities, rather than nefarious hiding of important study data.
One point to recognise is these analyses (there are several sub-sites running [1]) explore reporting of the study results onto specific public-domain databases, per FDAAA (or other) requirements. However, it's possible that results may be reported via other routes (poster or presentations at congresses, or publications in journals) - and this might be driven more by positive results than negative. I don't think the study methodology would catch examples which had been publicly reported, but via the wrong route. (These cases would still obviously be non-compliant, but maybe less of a shade of grey?)
From Ben Goldacre's paper[0] pharma industry sponsors are more likely to be compliant than non-industry sponsors, and sponsors of large trials more likely to be compliant than sponsors of small trials. This may suggest that it's partly a matter of human resource to actually compile the data and report it onto the public websites in question. Here, I'd also consider Hanlon's razor - busy people with other priorities, rather than nefarious hiding of important study data.
One point to recognise is these analyses (there are several sub-sites running [1]) explore reporting of the study results onto specific public-domain databases, per FDAAA (or other) requirements. However, it's possible that results may be reported via other routes (poster or presentations at congresses, or publications in journals) - and this might be driven more by positive results than negative. I don't think the study methodology would catch examples which had been publicly reported, but via the wrong route. (These cases would still obviously be non-compliant, but maybe less of a shade of grey?)
[0] https://www.thelancet.com/journals/lancet/article/PIIS0140-6... [1] https://trialstracker.net