I agree with Phil!
It's been my experience that L1(only) vectors are actually BETTER (more accurate, less noise) than L1/L2 vectors on short baselines (less than 10-20 kilometers depending on the conditions unique to a given set of observations).
My “rule of thumb” is 15 kilometers, but I have seen EXCELLENT L1(only) results out to 20 kilometers (and more). IT DEPENDS...
My general modus operandi is to process ALL SHORT baselines (under 10 or 15 kilometers) using L1 ONLY, regardless of the equipment used to collect the observations.
This of course ASSUMES that the L1 vectors “FIX” and the occupations are of reasonable duration.
I have run dozens of “tests” in a valley in Northern Utah that sports three PBO stations (one of which is in the NGS CORS Network), and processed hundreds of vectors both within the valley proper (to the three PBOs using several baseline processors) as well as half a dozen or so NGS CORS situate 50-250 kilometers distant therefrom (using both OPUS and Grafnet). In virtually every case, the L1 ONLY vector solution into the “local” PBO/CORS site was BETTER (actually quite a bit BETTER) than the L1/L2 vector solution to that site (even the OPUS/PAGES solution). I have heard that PAGES is “optimized” for “long vectors,” and in my experience, it (OPUS) doesn't do quite so good on SHORT (less than 10 kilometer) vectors (but I digress).
FLOAT solutions are another matter altogether.