Analyzing pool latency and rejects on L3+ (Scrypt vs other algos)

I've been doing some testing on my local network to see why my hardware efficiency is fluctuating so much. I noticed that a lot of people focus strictly on the 1% or 2% fee, but my logs show that network latency and stale shares are actually the bigger bottleneck for my setup.

For those of you still running older hardware like the L3+, have you noticed a significant difference in rejected share percentages between the major Scrypt pools and these newer "alternative" project pools?

I'm trying to figure out if switching is actually optimizing my hashrate stability or if I'm just seeing fluctuations due to local network jitter. What’s your average reject rate look like lately?