In the 1980s, there were dozens, if not hundreds, of companies making PC clones. Some used very careful and detailed engineering practices which systematically ensured that they would abide various timing constraints throughout the entire allowable ranges of manufacturing variations, temperature, and other such factors, but signal timings in an a 4.77 MHz 8088-based PC were sufficiently forgiving that such care wasn't needed to produce a machine that would work reliably with most common expansion cards, and thus many designers of PC clones did not exercise such care.
Intel was a sufficiently important company that it could ask IBM and Compaq to agree that they would not make any design changes to the machines that would cause deviations from the spec, without clearing such changes with Intel first, in exchange for having Intel endorse those companies' PCs as being compatible with the Inboard, and IBM and Compaq used a sufficiently rigorous design methodology that they could ensure that things like board layout revisions would not cause such deviations. A product like the Inboard would be likely to work just as well on many of the other machines whose designers made no effort to distinguish themselves from the PC in any way other than price, but it would have been impractical for Intel to reliably distinguish machines that would be proven to work by design, those that would likely work but whose could not be proven, and those that would in fact work.
