A purity figure is a single number describing how much of a material is the substance named on the label. It is the number people compare when they compare suppliers, and it is the number that tells you least about whether a material will work in your process.
At 99.5% assay, half a percent of what you received is something else. The assay does not say what.
Where the remainder comes from
Impurities are not contamination in the careless sense. They are mostly the fingerprint of how the material was made.
Route residues. A substance can usually be synthesised by more than one route, and each leaves a different set of by-products behind. Two manufacturers producing the same compound to the same assay can deliver materials whose remaining half-percent has nothing in common.
Catalyst traces. Where a metal catalyst was used, some of it remains. At the parts-per-million level this is invisible to an assay figure and highly visible to a process that is sensitive to that metal — plating baths in particular.
Residual solvents. What the material was crystallised or washed in, at whatever level the drying step left behind.
Isomers and closely related substances. Compounds that are chemically adjacent to the target, often difficult to separate, and sometimes not separated at all because the specification does not require it.
Water. Hygroscopic materials pick it up in storage and handling. It is usually specified separately from assay, and it is usually the difference between two lots that otherwise look identical.
Why this shows up as "the new lot behaves differently"
A process is qualified against the material it was developed on — including the parts of that material nobody characterised. When the supply changes, the assay can be identical and the process can still shift, because what changed was in the fraction the specification never described.
This is why continuity is often worth more than an incremental improvement in purity. A nominally better material with a different impurity profile is, from a process point of view, a different material. It may well be an improvement. It still needs to be treated as a change.
What to do about it
Specify what actually constrains you. If your process is sensitive to a particular metal, put a limit on that metal rather than asking for a higher general assay. A trace-metal limit is a question a supplier can answer; "as pure as possible" is not.
Ask what the specification does not cover. A certificate reports the tests the specification calls for. Anything outside that list is unmeasured rather than absent, and knowing which is which tells you where your risk sits.
Say whether you need lot continuity. If a process is running well on a qualified supply, that is worth stating on a repeat order. It changes what should be supplied, and it is easier to arrange in advance than to reconstruct afterwards.
Keep a retention sample. When a lot behaves differently, the useful comparison is against the material that worked, not against a specification both of them met.
The purity figure is still worth reading
None of this makes assay useless. It is the right first filter, it is what compendial monographs are built on, and a material outside its assay range is out of specification whatever else is true of it. The point is narrower: assay is a necessary check and an insufficient description, and treating it as a complete one is what produces the surprise on lot three.