Printed magazines asking for donations (for example) might not have — at time of printing — all of the information they want from donors, such as the name to put on a receipt, and where to send it to. A payment request with a more flexible reference might provide a way to tell the client what information to prompt the user for.
This could be in the form of description/tag pairs (such as "Email address"/"e", which would prompt for an "Email address" and result in a reference containing something like "e:firstname.lastname@example.org"), or it could just include a description of each field to prompt for, and have a more flexible reference in the payment specification, containing a sequence of the responses to the prompts, in the same order.
In either case, each requested piece of information could have multiple alternative descriptions, each tagged with the language of that description (see, for example, RFC 5646). Also, in either case, the payment request should also contain a fixed string to be included at the start of the reference in the payment specification.
But perhaps all this would add too much complexity, relative to the benefit it provides.