No, this is not about accusations that the SEC favors Wall Street firms over individual investors or others in its rulings. The accusation is far more subtle, that the agency is much better at catching certain types of malfeasance than others, because it is has basically adopted a "Wall Street" definition of what constitutes malfeasance.
In its earlier, 1930s incarnation, the SEC was tasked with tracking disclosure abuses; that is that firms that were issuing primary or secondary stock offerings were doing so either while lying about material facts about the company, or failing to disclose such material facts (those, which if known, were likely to change the investor's investment posture). Thus, the cardinal sin was one of misinformation.
Around 1970, that changed. Trading volumes exploded, raising new enforcement concerns. Meanwhile, new reporting technology (faster typewriters, then printers), as well as new accounting regulations, ensured that most firms at least went through the motions, if not the spirit, of "full disclosure." More "paper" was received by the SEC than it could reasonably process, meaning that "paper" violations were rare, and if present, went largely unchallenged. The new "sin" was not providing enough information, but rather, acting on privileged information to the detriment of others.
Such concerns about "trading," first surfaced in the SEC's investigation of Warren Buffett's Berkshire Hathaway (BRK/A), which later took over 80% of Wesco Savings and Loan. Wesco was being pursued by an unwanted suitor, with Berkshire being a potential "white knight." When word leaked out of Berkshire's interest, Buffett bent over backward to refuse shares that had earlier been tendered to him, meaning that the owners could re-sell at a higher price on the news. But the SEC had problems with this because Buffett had refused shares at the prevailing market price, and was presumably supporting the stock to keep it out of the reach of the other bidder, even to his own presumed detriment.
When the SEC asked Buffett if he wanted a lower or higher price, and Buffett said higher, he was considered a "market manipulator" because his strange trading pattern had pushed up the price of the stock, away from the takeover bidder. When queried further, Buffett admitted to making an offer but the seller declined to sell, meaning that there was no bid on the table. Buffett did buy stock at the higher price to woo the sellers, and eventually won control in a negotiated transaction. (Gre... which is legal but unethical, works the other way; the company buys out the suitor at an above market price.)
A more egregious example was the SEC's investigation of David Einhorn, a short-seller who was a "panner" of a company called Allied Capital (NYSEARCA:ALD). Einhorn's "sin" was to make a number of impromptu remarks at a charity luncheon that crystalized others' doubts about the stock, initiating a wave of selling (both short and exited longs) that drove the stock down several points. Einhorn was later exonerated when it was shown that he did not benefit from his remarks, and was therefore not guilty of "manipulation."
On the other hand, tracking a Bernie Maddoff scandal is beyond today's SEC. "No one," in their minds, plots a Ponzi scheme over a period of years and conducts it over a period of decades, covering their tracks with a false paper trail. "Record-keeping&q... violations are hardly part of the lexicon. "Everyone" on Wall street who runs a scam does so through improper trading of one sort or another. That's why Martha Stewart was prosecuted for a basically "victimless" crime.
But of the two violatioon, recording keeping" is almost certainly the more dangerous, because it harms many others by distorting their decisions, while trading violations often "only" result in the unjust enrichment of some (and unfair losses for their counterparties). It may not be so detrimental to the general public. But then again it may, because the investing public now relies on "technical" analysis (the record of trades) as much as or more than "fundamental" analysis of written records.
Disclosure: Long BRK/A, BRK/B, formerly long ALD.